1 /* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License (the "License"). 6 * You may not use this file except in compliance with the License. 7 * 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9 * or http://www.opensolaris.org/os/licensing. 10 * See the License for the specific language governing permissions 11 * and limitations under the License. 12 * 13 * When distributing Covered Code, include this CDDL HEADER in each 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15 * If applicable, add the following below this CDDL HEADER, with the 16 * fields enclosed by brackets "[]" replaced with your own identifying 17 * information: Portions Copyright [yyyy] [name of copyright owner] 18 * 19 * CDDL HEADER END 20 */ 21 22 /* 23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. 24 * Copyright 2011 Nexenta Systems, Inc. All rights reserved. 25 * Copyright (c) 2012, 2014 by Delphix. All rights reserved. 26 * Copyright (c) 2013, Joyent, Inc. All rights reserved. 27 */ 28 29 #include <ctype.h> 30 #include <errno.h> 31 #include <devid.h> 32 #include <fcntl.h> 33 #include <libintl.h> 34 #include <stdio.h> 35 #include <stdlib.h> 36 #include <strings.h> 37 #include <unistd.h> 38 #include <libgen.h> 39 #include <sys/efi_partition.h> 40 #include <sys/vtoc.h> 41 #include <sys/zfs_ioctl.h> 42 #include <dlfcn.h> 43 44 #include "zfs_namecheck.h" 45 #include "zfs_prop.h" 46 #include "libzfs_impl.h" 47 #include "zfs_comutil.h" 48 #include "zfeature_common.h" 49 50 static int read_efi_label(nvlist_t *config, diskaddr_t *sb); 51 52 #define DISK_ROOT "/dev/dsk" 53 #define RDISK_ROOT "/dev/rdsk" 54 #define BACKUP_SLICE "s2" 55 56 typedef struct prop_flags { 57 int create:1; /* Validate property on creation */ 58 int import:1; /* Validate property on import */ 59 } prop_flags_t; 60 61 /* 62 * ==================================================================== 63 * zpool property functions 64 * ==================================================================== 65 */ 66 67 static int 68 zpool_get_all_props(zpool_handle_t *zhp) 69 { 70 zfs_cmd_t zc = { 0 }; 71 libzfs_handle_t *hdl = zhp->zpool_hdl; 72 73 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 74 75 if (zcmd_alloc_dst_nvlist(hdl, &zc, 0) != 0) 76 return (-1); 77 78 while (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_PROPS, &zc) != 0) { 79 if (errno == ENOMEM) { 80 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) { 81 zcmd_free_nvlists(&zc); 82 return (-1); 83 } 84 } else { 85 zcmd_free_nvlists(&zc); 86 return (-1); 87 } 88 } 89 90 if (zcmd_read_dst_nvlist(hdl, &zc, &zhp->zpool_props) != 0) { 91 zcmd_free_nvlists(&zc); 92 return (-1); 93 } 94 95 zcmd_free_nvlists(&zc); 96 97 return (0); 98 } 99 100 static int 101 zpool_props_refresh(zpool_handle_t *zhp) 102 { 103 nvlist_t *old_props; 104 105 old_props = zhp->zpool_props; 106 107 if (zpool_get_all_props(zhp) != 0) 108 return (-1); 109 110 nvlist_free(old_props); 111 return (0); 112 } 113 114 static char * 115 zpool_get_prop_string(zpool_handle_t *zhp, zpool_prop_t prop, 116 zprop_source_t *src) 117 { 118 nvlist_t *nv, *nvl; 119 uint64_t ival; 120 char *value; 121 zprop_source_t source; 122 123 nvl = zhp->zpool_props; 124 if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) { 125 verify(nvlist_lookup_uint64(nv, ZPROP_SOURCE, &ival) == 0); 126 source = ival; 127 verify(nvlist_lookup_string(nv, ZPROP_VALUE, &value) == 0); 128 } else { 129 source = ZPROP_SRC_DEFAULT; 130 if ((value = (char *)zpool_prop_default_string(prop)) == NULL) 131 value = "-"; 132 } 133 134 if (src) 135 *src = source; 136 137 return (value); 138 } 139 140 uint64_t 141 zpool_get_prop_int(zpool_handle_t *zhp, zpool_prop_t prop, zprop_source_t *src) 142 { 143 nvlist_t *nv, *nvl; 144 uint64_t value; 145 zprop_source_t source; 146 147 if (zhp->zpool_props == NULL && zpool_get_all_props(zhp)) { 148 /* 149 * zpool_get_all_props() has most likely failed because 150 * the pool is faulted, but if all we need is the top level 151 * vdev's guid then get it from the zhp config nvlist. 152 */ 153 if ((prop == ZPOOL_PROP_GUID) && 154 (nvlist_lookup_nvlist(zhp->zpool_config, 155 ZPOOL_CONFIG_VDEV_TREE, &nv) == 0) && 156 (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &value) 157 == 0)) { 158 return (value); 159 } 160 return (zpool_prop_default_numeric(prop)); 161 } 162 163 nvl = zhp->zpool_props; 164 if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) { 165 verify(nvlist_lookup_uint64(nv, ZPROP_SOURCE, &value) == 0); 166 source = value; 167 verify(nvlist_lookup_uint64(nv, ZPROP_VALUE, &value) == 0); 168 } else { 169 source = ZPROP_SRC_DEFAULT; 170 value = zpool_prop_default_numeric(prop); 171 } 172 173 if (src) 174 *src = source; 175 176 return (value); 177 } 178 179 /* 180 * Map VDEV STATE to printed strings. 181 */ 182 char * 183 zpool_state_to_name(vdev_state_t state, vdev_aux_t aux) 184 { 185 switch (state) { 186 case VDEV_STATE_CLOSED: 187 case VDEV_STATE_OFFLINE: 188 return (gettext("OFFLINE")); 189 case VDEV_STATE_REMOVED: 190 return (gettext("REMOVED")); 191 case VDEV_STATE_CANT_OPEN: 192 if (aux == VDEV_AUX_CORRUPT_DATA || aux == VDEV_AUX_BAD_LOG) 193 return (gettext("FAULTED")); 194 else if (aux == VDEV_AUX_SPLIT_POOL) 195 return (gettext("SPLIT")); 196 else 197 return (gettext("UNAVAIL")); 198 case VDEV_STATE_FAULTED: 199 return (gettext("FAULTED")); 200 case VDEV_STATE_DEGRADED: 201 return (gettext("DEGRADED")); 202 case VDEV_STATE_HEALTHY: 203 return (gettext("ONLINE")); 204 } 205 206 return (gettext("UNKNOWN")); 207 } 208 209 /* 210 * Get a zpool property value for 'prop' and return the value in 211 * a pre-allocated buffer. 212 */ 213 int 214 zpool_get_prop(zpool_handle_t *zhp, zpool_prop_t prop, char *buf, size_t len, 215 zprop_source_t *srctype, boolean_t literal) 216 { 217 uint64_t intval; 218 const char *strval; 219 zprop_source_t src = ZPROP_SRC_NONE; 220 nvlist_t *nvroot; 221 vdev_stat_t *vs; 222 uint_t vsc; 223 224 if (zpool_get_state(zhp) == POOL_STATE_UNAVAIL) { 225 switch (prop) { 226 case ZPOOL_PROP_NAME: 227 (void) strlcpy(buf, zpool_get_name(zhp), len); 228 break; 229 230 case ZPOOL_PROP_HEALTH: 231 (void) strlcpy(buf, "FAULTED", len); 232 break; 233 234 case ZPOOL_PROP_GUID: 235 intval = zpool_get_prop_int(zhp, prop, &src); 236 (void) snprintf(buf, len, "%llu", intval); 237 break; 238 239 case ZPOOL_PROP_ALTROOT: 240 case ZPOOL_PROP_CACHEFILE: 241 case ZPOOL_PROP_COMMENT: 242 if (zhp->zpool_props != NULL || 243 zpool_get_all_props(zhp) == 0) { 244 (void) strlcpy(buf, 245 zpool_get_prop_string(zhp, prop, &src), 246 len); 247 break; 248 } 249 /* FALLTHROUGH */ 250 default: 251 (void) strlcpy(buf, "-", len); 252 break; 253 } 254 255 if (srctype != NULL) 256 *srctype = src; 257 return (0); 258 } 259 260 if (zhp->zpool_props == NULL && zpool_get_all_props(zhp) && 261 prop != ZPOOL_PROP_NAME) 262 return (-1); 263 264 switch (zpool_prop_get_type(prop)) { 265 case PROP_TYPE_STRING: 266 (void) strlcpy(buf, zpool_get_prop_string(zhp, prop, &src), 267 len); 268 break; 269 270 case PROP_TYPE_NUMBER: 271 intval = zpool_get_prop_int(zhp, prop, &src); 272 273 switch (prop) { 274 case ZPOOL_PROP_SIZE: 275 case ZPOOL_PROP_ALLOCATED: 276 case ZPOOL_PROP_FREE: 277 case ZPOOL_PROP_FREEING: 278 case ZPOOL_PROP_LEAKED: 279 case ZPOOL_PROP_EXPANDSZ: 280 if (literal) { 281 (void) snprintf(buf, len, "%llu", 282 (u_longlong_t)intval); 283 } else { 284 (void) zfs_nicenum(intval, buf, len); 285 } 286 break; 287 288 case ZPOOL_PROP_CAPACITY: 289 if (literal) { 290 (void) snprintf(buf, len, "%llu", 291 (u_longlong_t)intval); 292 } else { 293 (void) snprintf(buf, len, "%llu%%", 294 (u_longlong_t)intval); 295 } 296 break; 297 298 case ZPOOL_PROP_DEDUPRATIO: 299 (void) snprintf(buf, len, "%llu.%02llux", 300 (u_longlong_t)(intval / 100), 301 (u_longlong_t)(intval % 100)); 302 break; 303 304 case ZPOOL_PROP_HEALTH: 305 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL), 306 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); 307 verify(nvlist_lookup_uint64_array(nvroot, 308 ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs, &vsc) 309 == 0); 310 311 (void) strlcpy(buf, zpool_state_to_name(intval, 312 vs->vs_aux), len); 313 break; 314 case ZPOOL_PROP_VERSION: 315 if (intval >= SPA_VERSION_FEATURES) { 316 (void) snprintf(buf, len, "-"); 317 break; 318 } 319 /* FALLTHROUGH */ 320 default: 321 (void) snprintf(buf, len, "%llu", intval); 322 } 323 break; 324 325 case PROP_TYPE_INDEX: 326 intval = zpool_get_prop_int(zhp, prop, &src); 327 if (zpool_prop_index_to_string(prop, intval, &strval) 328 != 0) 329 return (-1); 330 (void) strlcpy(buf, strval, len); 331 break; 332 333 default: 334 abort(); 335 } 336 337 if (srctype) 338 *srctype = src; 339 340 return (0); 341 } 342 343 /* 344 * Check if the bootfs name has the same pool name as it is set to. 345 * Assuming bootfs is a valid dataset name. 346 */ 347 static boolean_t 348 bootfs_name_valid(const char *pool, char *bootfs) 349 { 350 int len = strlen(pool); 351 352 if (!zfs_name_valid(bootfs, ZFS_TYPE_FILESYSTEM|ZFS_TYPE_SNAPSHOT)) 353 return (B_FALSE); 354 355 if (strncmp(pool, bootfs, len) == 0 && 356 (bootfs[len] == '/' || bootfs[len] == '\0')) 357 return (B_TRUE); 358 359 return (B_FALSE); 360 } 361 362 /* 363 * Inspect the configuration to determine if any of the devices contain 364 * an EFI label. 365 */ 366 static boolean_t 367 pool_uses_efi(nvlist_t *config) 368 { 369 nvlist_t **child; 370 uint_t c, children; 371 372 if (nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_CHILDREN, 373 &child, &children) != 0) 374 return (read_efi_label(config, NULL) >= 0); 375 376 for (c = 0; c < children; c++) { 377 if (pool_uses_efi(child[c])) 378 return (B_TRUE); 379 } 380 return (B_FALSE); 381 } 382 383 boolean_t 384 zpool_is_bootable(zpool_handle_t *zhp) 385 { 386 char bootfs[ZPOOL_MAXNAMELEN]; 387 388 return (zpool_get_prop(zhp, ZPOOL_PROP_BOOTFS, bootfs, 389 sizeof (bootfs), NULL, B_FALSE) == 0 && strncmp(bootfs, "-", 390 sizeof (bootfs)) != 0); 391 } 392 393 394 /* 395 * Given an nvlist of zpool properties to be set, validate that they are 396 * correct, and parse any numeric properties (index, boolean, etc) if they are 397 * specified as strings. 398 */ 399 static nvlist_t * 400 zpool_valid_proplist(libzfs_handle_t *hdl, const char *poolname, 401 nvlist_t *props, uint64_t version, prop_flags_t flags, char *errbuf) 402 { 403 nvpair_t *elem; 404 nvlist_t *retprops; 405 zpool_prop_t prop; 406 char *strval; 407 uint64_t intval; 408 char *slash, *check; 409 struct stat64 statbuf; 410 zpool_handle_t *zhp; 411 nvlist_t *nvroot; 412 413 if (nvlist_alloc(&retprops, NV_UNIQUE_NAME, 0) != 0) { 414 (void) no_memory(hdl); 415 return (NULL); 416 } 417 418 elem = NULL; 419 while ((elem = nvlist_next_nvpair(props, elem)) != NULL) { 420 const char *propname = nvpair_name(elem); 421 422 prop = zpool_name_to_prop(propname); 423 if (prop == ZPROP_INVAL && zpool_prop_feature(propname)) { 424 int err; 425 char *fname = strchr(propname, '@') + 1; 426 427 err = zfeature_lookup_name(fname, NULL); 428 if (err != 0) { 429 ASSERT3U(err, ==, ENOENT); 430 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 431 "invalid feature '%s'"), fname); 432 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 433 goto error; 434 } 435 436 if (nvpair_type(elem) != DATA_TYPE_STRING) { 437 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 438 "'%s' must be a string"), propname); 439 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 440 goto error; 441 } 442 443 (void) nvpair_value_string(elem, &strval); 444 if (strcmp(strval, ZFS_FEATURE_ENABLED) != 0) { 445 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 446 "property '%s' can only be set to " 447 "'enabled'"), propname); 448 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 449 goto error; 450 } 451 452 if (nvlist_add_uint64(retprops, propname, 0) != 0) { 453 (void) no_memory(hdl); 454 goto error; 455 } 456 continue; 457 } 458 459 /* 460 * Make sure this property is valid and applies to this type. 461 */ 462 if (prop == ZPROP_INVAL) { 463 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 464 "invalid property '%s'"), propname); 465 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 466 goto error; 467 } 468 469 if (zpool_prop_readonly(prop)) { 470 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' " 471 "is readonly"), propname); 472 (void) zfs_error(hdl, EZFS_PROPREADONLY, errbuf); 473 goto error; 474 } 475 476 if (zprop_parse_value(hdl, elem, prop, ZFS_TYPE_POOL, retprops, 477 &strval, &intval, errbuf) != 0) 478 goto error; 479 480 /* 481 * Perform additional checking for specific properties. 482 */ 483 switch (prop) { 484 case ZPOOL_PROP_VERSION: 485 if (intval < version || 486 !SPA_VERSION_IS_SUPPORTED(intval)) { 487 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 488 "property '%s' number %d is invalid."), 489 propname, intval); 490 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf); 491 goto error; 492 } 493 break; 494 495 case ZPOOL_PROP_BOOTFS: 496 if (flags.create || flags.import) { 497 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 498 "property '%s' cannot be set at creation " 499 "or import time"), propname); 500 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 501 goto error; 502 } 503 504 if (version < SPA_VERSION_BOOTFS) { 505 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 506 "pool must be upgraded to support " 507 "'%s' property"), propname); 508 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf); 509 goto error; 510 } 511 512 /* 513 * bootfs property value has to be a dataset name and 514 * the dataset has to be in the same pool as it sets to. 515 */ 516 if (strval[0] != '\0' && !bootfs_name_valid(poolname, 517 strval)) { 518 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' " 519 "is an invalid name"), strval); 520 (void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf); 521 goto error; 522 } 523 524 if ((zhp = zpool_open_canfail(hdl, poolname)) == NULL) { 525 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 526 "could not open pool '%s'"), poolname); 527 (void) zfs_error(hdl, EZFS_OPENFAILED, errbuf); 528 goto error; 529 } 530 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL), 531 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); 532 533 /* 534 * bootfs property cannot be set on a disk which has 535 * been EFI labeled. 536 */ 537 if (pool_uses_efi(nvroot)) { 538 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 539 "property '%s' not supported on " 540 "EFI labeled devices"), propname); 541 (void) zfs_error(hdl, EZFS_POOL_NOTSUP, errbuf); 542 zpool_close(zhp); 543 goto error; 544 } 545 zpool_close(zhp); 546 break; 547 548 case ZPOOL_PROP_ALTROOT: 549 if (!flags.create && !flags.import) { 550 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 551 "property '%s' can only be set during pool " 552 "creation or import"), propname); 553 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 554 goto error; 555 } 556 557 if (strval[0] != '/') { 558 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 559 "bad alternate root '%s'"), strval); 560 (void) zfs_error(hdl, EZFS_BADPATH, errbuf); 561 goto error; 562 } 563 break; 564 565 case ZPOOL_PROP_CACHEFILE: 566 if (strval[0] == '\0') 567 break; 568 569 if (strcmp(strval, "none") == 0) 570 break; 571 572 if (strval[0] != '/') { 573 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 574 "property '%s' must be empty, an " 575 "absolute path, or 'none'"), propname); 576 (void) zfs_error(hdl, EZFS_BADPATH, errbuf); 577 goto error; 578 } 579 580 slash = strrchr(strval, '/'); 581 582 if (slash[1] == '\0' || strcmp(slash, "/.") == 0 || 583 strcmp(slash, "/..") == 0) { 584 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 585 "'%s' is not a valid file"), strval); 586 (void) zfs_error(hdl, EZFS_BADPATH, errbuf); 587 goto error; 588 } 589 590 *slash = '\0'; 591 592 if (strval[0] != '\0' && 593 (stat64(strval, &statbuf) != 0 || 594 !S_ISDIR(statbuf.st_mode))) { 595 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 596 "'%s' is not a valid directory"), 597 strval); 598 (void) zfs_error(hdl, EZFS_BADPATH, errbuf); 599 goto error; 600 } 601 602 *slash = '/'; 603 break; 604 605 case ZPOOL_PROP_COMMENT: 606 for (check = strval; *check != '\0'; check++) { 607 if (!isprint(*check)) { 608 zfs_error_aux(hdl, 609 dgettext(TEXT_DOMAIN, 610 "comment may only have printable " 611 "characters")); 612 (void) zfs_error(hdl, EZFS_BADPROP, 613 errbuf); 614 goto error; 615 } 616 } 617 if (strlen(strval) > ZPROP_MAX_COMMENT) { 618 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 619 "comment must not exceed %d characters"), 620 ZPROP_MAX_COMMENT); 621 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 622 goto error; 623 } 624 break; 625 case ZPOOL_PROP_READONLY: 626 if (!flags.import) { 627 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 628 "property '%s' can only be set at " 629 "import time"), propname); 630 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 631 goto error; 632 } 633 break; 634 } 635 } 636 637 return (retprops); 638 error: 639 nvlist_free(retprops); 640 return (NULL); 641 } 642 643 /* 644 * Set zpool property : propname=propval. 645 */ 646 int 647 zpool_set_prop(zpool_handle_t *zhp, const char *propname, const char *propval) 648 { 649 zfs_cmd_t zc = { 0 }; 650 int ret = -1; 651 char errbuf[1024]; 652 nvlist_t *nvl = NULL; 653 nvlist_t *realprops; 654 uint64_t version; 655 prop_flags_t flags = { 0 }; 656 657 (void) snprintf(errbuf, sizeof (errbuf), 658 dgettext(TEXT_DOMAIN, "cannot set property for '%s'"), 659 zhp->zpool_name); 660 661 if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0) 662 return (no_memory(zhp->zpool_hdl)); 663 664 if (nvlist_add_string(nvl, propname, propval) != 0) { 665 nvlist_free(nvl); 666 return (no_memory(zhp->zpool_hdl)); 667 } 668 669 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL); 670 if ((realprops = zpool_valid_proplist(zhp->zpool_hdl, 671 zhp->zpool_name, nvl, version, flags, errbuf)) == NULL) { 672 nvlist_free(nvl); 673 return (-1); 674 } 675 676 nvlist_free(nvl); 677 nvl = realprops; 678 679 /* 680 * Execute the corresponding ioctl() to set this property. 681 */ 682 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 683 684 if (zcmd_write_src_nvlist(zhp->zpool_hdl, &zc, nvl) != 0) { 685 nvlist_free(nvl); 686 return (-1); 687 } 688 689 ret = zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_SET_PROPS, &zc); 690 691 zcmd_free_nvlists(&zc); 692 nvlist_free(nvl); 693 694 if (ret) 695 (void) zpool_standard_error(zhp->zpool_hdl, errno, errbuf); 696 else 697 (void) zpool_props_refresh(zhp); 698 699 return (ret); 700 } 701 702 int 703 zpool_expand_proplist(zpool_handle_t *zhp, zprop_list_t **plp) 704 { 705 libzfs_handle_t *hdl = zhp->zpool_hdl; 706 zprop_list_t *entry; 707 char buf[ZFS_MAXPROPLEN]; 708 nvlist_t *features = NULL; 709 zprop_list_t **last; 710 boolean_t firstexpand = (NULL == *plp); 711 712 if (zprop_expand_list(hdl, plp, ZFS_TYPE_POOL) != 0) 713 return (-1); 714 715 last = plp; 716 while (*last != NULL) 717 last = &(*last)->pl_next; 718 719 if ((*plp)->pl_all) 720 features = zpool_get_features(zhp); 721 722 if ((*plp)->pl_all && firstexpand) { 723 for (int i = 0; i < SPA_FEATURES; i++) { 724 zprop_list_t *entry = zfs_alloc(hdl, 725 sizeof (zprop_list_t)); 726 entry->pl_prop = ZPROP_INVAL; 727 entry->pl_user_prop = zfs_asprintf(hdl, "feature@%s", 728 spa_feature_table[i].fi_uname); 729 entry->pl_width = strlen(entry->pl_user_prop); 730 entry->pl_all = B_TRUE; 731 732 *last = entry; 733 last = &entry->pl_next; 734 } 735 } 736 737 /* add any unsupported features */ 738 for (nvpair_t *nvp = nvlist_next_nvpair(features, NULL); 739 nvp != NULL; nvp = nvlist_next_nvpair(features, nvp)) { 740 char *propname; 741 boolean_t found; 742 zprop_list_t *entry; 743 744 if (zfeature_is_supported(nvpair_name(nvp))) 745 continue; 746 747 propname = zfs_asprintf(hdl, "unsupported@%s", 748 nvpair_name(nvp)); 749 750 /* 751 * Before adding the property to the list make sure that no 752 * other pool already added the same property. 753 */ 754 found = B_FALSE; 755 entry = *plp; 756 while (entry != NULL) { 757 if (entry->pl_user_prop != NULL && 758 strcmp(propname, entry->pl_user_prop) == 0) { 759 found = B_TRUE; 760 break; 761 } 762 entry = entry->pl_next; 763 } 764 if (found) { 765 free(propname); 766 continue; 767 } 768 769 entry = zfs_alloc(hdl, sizeof (zprop_list_t)); 770 entry->pl_prop = ZPROP_INVAL; 771 entry->pl_user_prop = propname; 772 entry->pl_width = strlen(entry->pl_user_prop); 773 entry->pl_all = B_TRUE; 774 775 *last = entry; 776 last = &entry->pl_next; 777 } 778 779 for (entry = *plp; entry != NULL; entry = entry->pl_next) { 780 781 if (entry->pl_fixed) 782 continue; 783 784 if (entry->pl_prop != ZPROP_INVAL && 785 zpool_get_prop(zhp, entry->pl_prop, buf, sizeof (buf), 786 NULL, B_FALSE) == 0) { 787 if (strlen(buf) > entry->pl_width) 788 entry->pl_width = strlen(buf); 789 } 790 } 791 792 return (0); 793 } 794 795 /* 796 * Get the state for the given feature on the given ZFS pool. 797 */ 798 int 799 zpool_prop_get_feature(zpool_handle_t *zhp, const char *propname, char *buf, 800 size_t len) 801 { 802 uint64_t refcount; 803 boolean_t found = B_FALSE; 804 nvlist_t *features = zpool_get_features(zhp); 805 boolean_t supported; 806 const char *feature = strchr(propname, '@') + 1; 807 808 supported = zpool_prop_feature(propname); 809 ASSERT(supported || zfs_prop_unsupported(propname)); 810 811 /* 812 * Convert from feature name to feature guid. This conversion is 813 * unecessary for unsupported@... properties because they already 814 * use guids. 815 */ 816 if (supported) { 817 int ret; 818 spa_feature_t fid; 819 820 ret = zfeature_lookup_name(feature, &fid); 821 if (ret != 0) { 822 (void) strlcpy(buf, "-", len); 823 return (ENOTSUP); 824 } 825 feature = spa_feature_table[fid].fi_guid; 826 } 827 828 if (nvlist_lookup_uint64(features, feature, &refcount) == 0) 829 found = B_TRUE; 830 831 if (supported) { 832 if (!found) { 833 (void) strlcpy(buf, ZFS_FEATURE_DISABLED, len); 834 } else { 835 if (refcount == 0) 836 (void) strlcpy(buf, ZFS_FEATURE_ENABLED, len); 837 else 838 (void) strlcpy(buf, ZFS_FEATURE_ACTIVE, len); 839 } 840 } else { 841 if (found) { 842 if (refcount == 0) { 843 (void) strcpy(buf, ZFS_UNSUPPORTED_INACTIVE); 844 } else { 845 (void) strcpy(buf, ZFS_UNSUPPORTED_READONLY); 846 } 847 } else { 848 (void) strlcpy(buf, "-", len); 849 return (ENOTSUP); 850 } 851 } 852 853 return (0); 854 } 855 856 /* 857 * Don't start the slice at the default block of 34; many storage 858 * devices will use a stripe width of 128k, so start there instead. 859 */ 860 #define NEW_START_BLOCK 256 861 862 /* 863 * Validate the given pool name, optionally putting an extended error message in 864 * 'buf'. 865 */ 866 boolean_t 867 zpool_name_valid(libzfs_handle_t *hdl, boolean_t isopen, const char *pool) 868 { 869 namecheck_err_t why; 870 char what; 871 int ret; 872 873 ret = pool_namecheck(pool, &why, &what); 874 875 /* 876 * The rules for reserved pool names were extended at a later point. 877 * But we need to support users with existing pools that may now be 878 * invalid. So we only check for this expanded set of names during a 879 * create (or import), and only in userland. 880 */ 881 if (ret == 0 && !isopen && 882 (strncmp(pool, "mirror", 6) == 0 || 883 strncmp(pool, "raidz", 5) == 0 || 884 strncmp(pool, "spare", 5) == 0 || 885 strcmp(pool, "log") == 0)) { 886 if (hdl != NULL) 887 zfs_error_aux(hdl, 888 dgettext(TEXT_DOMAIN, "name is reserved")); 889 return (B_FALSE); 890 } 891 892 893 if (ret != 0) { 894 if (hdl != NULL) { 895 switch (why) { 896 case NAME_ERR_TOOLONG: 897 zfs_error_aux(hdl, 898 dgettext(TEXT_DOMAIN, "name is too long")); 899 break; 900 901 case NAME_ERR_INVALCHAR: 902 zfs_error_aux(hdl, 903 dgettext(TEXT_DOMAIN, "invalid character " 904 "'%c' in pool name"), what); 905 break; 906 907 case NAME_ERR_NOLETTER: 908 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 909 "name must begin with a letter")); 910 break; 911 912 case NAME_ERR_RESERVED: 913 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 914 "name is reserved")); 915 break; 916 917 case NAME_ERR_DISKLIKE: 918 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 919 "pool name is reserved")); 920 break; 921 922 case NAME_ERR_LEADING_SLASH: 923 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 924 "leading slash in name")); 925 break; 926 927 case NAME_ERR_EMPTY_COMPONENT: 928 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 929 "empty component in name")); 930 break; 931 932 case NAME_ERR_TRAILING_SLASH: 933 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 934 "trailing slash in name")); 935 break; 936 937 case NAME_ERR_MULTIPLE_AT: 938 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 939 "multiple '@' delimiters in name")); 940 break; 941 942 } 943 } 944 return (B_FALSE); 945 } 946 947 return (B_TRUE); 948 } 949 950 /* 951 * Open a handle to the given pool, even if the pool is currently in the FAULTED 952 * state. 953 */ 954 zpool_handle_t * 955 zpool_open_canfail(libzfs_handle_t *hdl, const char *pool) 956 { 957 zpool_handle_t *zhp; 958 boolean_t missing; 959 960 /* 961 * Make sure the pool name is valid. 962 */ 963 if (!zpool_name_valid(hdl, B_TRUE, pool)) { 964 (void) zfs_error_fmt(hdl, EZFS_INVALIDNAME, 965 dgettext(TEXT_DOMAIN, "cannot open '%s'"), 966 pool); 967 return (NULL); 968 } 969 970 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL) 971 return (NULL); 972 973 zhp->zpool_hdl = hdl; 974 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name)); 975 976 if (zpool_refresh_stats(zhp, &missing) != 0) { 977 zpool_close(zhp); 978 return (NULL); 979 } 980 981 if (missing) { 982 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "no such pool")); 983 (void) zfs_error_fmt(hdl, EZFS_NOENT, 984 dgettext(TEXT_DOMAIN, "cannot open '%s'"), pool); 985 zpool_close(zhp); 986 return (NULL); 987 } 988 989 return (zhp); 990 } 991 992 /* 993 * Like the above, but silent on error. Used when iterating over pools (because 994 * the configuration cache may be out of date). 995 */ 996 int 997 zpool_open_silent(libzfs_handle_t *hdl, const char *pool, zpool_handle_t **ret) 998 { 999 zpool_handle_t *zhp; 1000 boolean_t missing; 1001 1002 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL) 1003 return (-1); 1004 1005 zhp->zpool_hdl = hdl; 1006 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name)); 1007 1008 if (zpool_refresh_stats(zhp, &missing) != 0) { 1009 zpool_close(zhp); 1010 return (-1); 1011 } 1012 1013 if (missing) { 1014 zpool_close(zhp); 1015 *ret = NULL; 1016 return (0); 1017 } 1018 1019 *ret = zhp; 1020 return (0); 1021 } 1022 1023 /* 1024 * Similar to zpool_open_canfail(), but refuses to open pools in the faulted 1025 * state. 1026 */ 1027 zpool_handle_t * 1028 zpool_open(libzfs_handle_t *hdl, const char *pool) 1029 { 1030 zpool_handle_t *zhp; 1031 1032 if ((zhp = zpool_open_canfail(hdl, pool)) == NULL) 1033 return (NULL); 1034 1035 if (zhp->zpool_state == POOL_STATE_UNAVAIL) { 1036 (void) zfs_error_fmt(hdl, EZFS_POOLUNAVAIL, 1037 dgettext(TEXT_DOMAIN, "cannot open '%s'"), zhp->zpool_name); 1038 zpool_close(zhp); 1039 return (NULL); 1040 } 1041 1042 return (zhp); 1043 } 1044 1045 /* 1046 * Close the handle. Simply frees the memory associated with the handle. 1047 */ 1048 void 1049 zpool_close(zpool_handle_t *zhp) 1050 { 1051 if (zhp->zpool_config) 1052 nvlist_free(zhp->zpool_config); 1053 if (zhp->zpool_old_config) 1054 nvlist_free(zhp->zpool_old_config); 1055 if (zhp->zpool_props) 1056 nvlist_free(zhp->zpool_props); 1057 free(zhp); 1058 } 1059 1060 /* 1061 * Return the name of the pool. 1062 */ 1063 const char * 1064 zpool_get_name(zpool_handle_t *zhp) 1065 { 1066 return (zhp->zpool_name); 1067 } 1068 1069 1070 /* 1071 * Return the state of the pool (ACTIVE or UNAVAILABLE) 1072 */ 1073 int 1074 zpool_get_state(zpool_handle_t *zhp) 1075 { 1076 return (zhp->zpool_state); 1077 } 1078 1079 /* 1080 * Create the named pool, using the provided vdev list. It is assumed 1081 * that the consumer has already validated the contents of the nvlist, so we 1082 * don't have to worry about error semantics. 1083 */ 1084 int 1085 zpool_create(libzfs_handle_t *hdl, const char *pool, nvlist_t *nvroot, 1086 nvlist_t *props, nvlist_t *fsprops) 1087 { 1088 zfs_cmd_t zc = { 0 }; 1089 nvlist_t *zc_fsprops = NULL; 1090 nvlist_t *zc_props = NULL; 1091 char msg[1024]; 1092 int ret = -1; 1093 1094 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 1095 "cannot create '%s'"), pool); 1096 1097 if (!zpool_name_valid(hdl, B_FALSE, pool)) 1098 return (zfs_error(hdl, EZFS_INVALIDNAME, msg)); 1099 1100 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0) 1101 return (-1); 1102 1103 if (props) { 1104 prop_flags_t flags = { .create = B_TRUE, .import = B_FALSE }; 1105 1106 if ((zc_props = zpool_valid_proplist(hdl, pool, props, 1107 SPA_VERSION_1, flags, msg)) == NULL) { 1108 goto create_failed; 1109 } 1110 } 1111 1112 if (fsprops) { 1113 uint64_t zoned; 1114 char *zonestr; 1115 1116 zoned = ((nvlist_lookup_string(fsprops, 1117 zfs_prop_to_name(ZFS_PROP_ZONED), &zonestr) == 0) && 1118 strcmp(zonestr, "on") == 0); 1119 1120 if ((zc_fsprops = zfs_valid_proplist(hdl, 1121 ZFS_TYPE_FILESYSTEM, fsprops, zoned, NULL, msg)) == NULL) { 1122 goto create_failed; 1123 } 1124 if (!zc_props && 1125 (nvlist_alloc(&zc_props, NV_UNIQUE_NAME, 0) != 0)) { 1126 goto create_failed; 1127 } 1128 if (nvlist_add_nvlist(zc_props, 1129 ZPOOL_ROOTFS_PROPS, zc_fsprops) != 0) { 1130 goto create_failed; 1131 } 1132 } 1133 1134 if (zc_props && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0) 1135 goto create_failed; 1136 1137 (void) strlcpy(zc.zc_name, pool, sizeof (zc.zc_name)); 1138 1139 if ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_CREATE, &zc)) != 0) { 1140 1141 zcmd_free_nvlists(&zc); 1142 nvlist_free(zc_props); 1143 nvlist_free(zc_fsprops); 1144 1145 switch (errno) { 1146 case EBUSY: 1147 /* 1148 * This can happen if the user has specified the same 1149 * device multiple times. We can't reliably detect this 1150 * until we try to add it and see we already have a 1151 * label. 1152 */ 1153 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1154 "one or more vdevs refer to the same device")); 1155 return (zfs_error(hdl, EZFS_BADDEV, msg)); 1156 1157 case EOVERFLOW: 1158 /* 1159 * This occurs when one of the devices is below 1160 * SPA_MINDEVSIZE. Unfortunately, we can't detect which 1161 * device was the problem device since there's no 1162 * reliable way to determine device size from userland. 1163 */ 1164 { 1165 char buf[64]; 1166 1167 zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf)); 1168 1169 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1170 "one or more devices is less than the " 1171 "minimum size (%s)"), buf); 1172 } 1173 return (zfs_error(hdl, EZFS_BADDEV, msg)); 1174 1175 case ENOSPC: 1176 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1177 "one or more devices is out of space")); 1178 return (zfs_error(hdl, EZFS_BADDEV, msg)); 1179 1180 case ENOTBLK: 1181 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1182 "cache device must be a disk or disk slice")); 1183 return (zfs_error(hdl, EZFS_BADDEV, msg)); 1184 1185 default: 1186 return (zpool_standard_error(hdl, errno, msg)); 1187 } 1188 } 1189 1190 create_failed: 1191 zcmd_free_nvlists(&zc); 1192 nvlist_free(zc_props); 1193 nvlist_free(zc_fsprops); 1194 return (ret); 1195 } 1196 1197 /* 1198 * Destroy the given pool. It is up to the caller to ensure that there are no 1199 * datasets left in the pool. 1200 */ 1201 int 1202 zpool_destroy(zpool_handle_t *zhp, const char *log_str) 1203 { 1204 zfs_cmd_t zc = { 0 }; 1205 zfs_handle_t *zfp = NULL; 1206 libzfs_handle_t *hdl = zhp->zpool_hdl; 1207 char msg[1024]; 1208 1209 if (zhp->zpool_state == POOL_STATE_ACTIVE && 1210 (zfp = zfs_open(hdl, zhp->zpool_name, ZFS_TYPE_FILESYSTEM)) == NULL) 1211 return (-1); 1212 1213 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 1214 zc.zc_history = (uint64_t)(uintptr_t)log_str; 1215 1216 if (zfs_ioctl(hdl, ZFS_IOC_POOL_DESTROY, &zc) != 0) { 1217 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 1218 "cannot destroy '%s'"), zhp->zpool_name); 1219 1220 if (errno == EROFS) { 1221 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1222 "one or more devices is read only")); 1223 (void) zfs_error(hdl, EZFS_BADDEV, msg); 1224 } else { 1225 (void) zpool_standard_error(hdl, errno, msg); 1226 } 1227 1228 if (zfp) 1229 zfs_close(zfp); 1230 return (-1); 1231 } 1232 1233 if (zfp) { 1234 remove_mountpoint(zfp); 1235 zfs_close(zfp); 1236 } 1237 1238 return (0); 1239 } 1240 1241 /* 1242 * Add the given vdevs to the pool. The caller must have already performed the 1243 * necessary verification to ensure that the vdev specification is well-formed. 1244 */ 1245 int 1246 zpool_add(zpool_handle_t *zhp, nvlist_t *nvroot) 1247 { 1248 zfs_cmd_t zc = { 0 }; 1249 int ret; 1250 libzfs_handle_t *hdl = zhp->zpool_hdl; 1251 char msg[1024]; 1252 nvlist_t **spares, **l2cache; 1253 uint_t nspares, nl2cache; 1254 1255 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 1256 "cannot add to '%s'"), zhp->zpool_name); 1257 1258 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) < 1259 SPA_VERSION_SPARES && 1260 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES, 1261 &spares, &nspares) == 0) { 1262 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be " 1263 "upgraded to add hot spares")); 1264 return (zfs_error(hdl, EZFS_BADVERSION, msg)); 1265 } 1266 1267 if (zpool_is_bootable(zhp) && nvlist_lookup_nvlist_array(nvroot, 1268 ZPOOL_CONFIG_SPARES, &spares, &nspares) == 0) { 1269 uint64_t s; 1270 1271 for (s = 0; s < nspares; s++) { 1272 char *path; 1273 1274 if (nvlist_lookup_string(spares[s], ZPOOL_CONFIG_PATH, 1275 &path) == 0 && pool_uses_efi(spares[s])) { 1276 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1277 "device '%s' contains an EFI label and " 1278 "cannot be used on root pools."), 1279 zpool_vdev_name(hdl, NULL, spares[s], 1280 B_FALSE)); 1281 return (zfs_error(hdl, EZFS_POOL_NOTSUP, msg)); 1282 } 1283 } 1284 } 1285 1286 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) < 1287 SPA_VERSION_L2CACHE && 1288 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE, 1289 &l2cache, &nl2cache) == 0) { 1290 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be " 1291 "upgraded to add cache devices")); 1292 return (zfs_error(hdl, EZFS_BADVERSION, msg)); 1293 } 1294 1295 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0) 1296 return (-1); 1297 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 1298 1299 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_ADD, &zc) != 0) { 1300 switch (errno) { 1301 case EBUSY: 1302 /* 1303 * This can happen if the user has specified the same 1304 * device multiple times. We can't reliably detect this 1305 * until we try to add it and see we already have a 1306 * label. 1307 */ 1308 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1309 "one or more vdevs refer to the same device")); 1310 (void) zfs_error(hdl, EZFS_BADDEV, msg); 1311 break; 1312 1313 case EOVERFLOW: 1314 /* 1315 * This occurrs when one of the devices is below 1316 * SPA_MINDEVSIZE. Unfortunately, we can't detect which 1317 * device was the problem device since there's no 1318 * reliable way to determine device size from userland. 1319 */ 1320 { 1321 char buf[64]; 1322 1323 zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf)); 1324 1325 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1326 "device is less than the minimum " 1327 "size (%s)"), buf); 1328 } 1329 (void) zfs_error(hdl, EZFS_BADDEV, msg); 1330 break; 1331 1332 case ENOTSUP: 1333 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1334 "pool must be upgraded to add these vdevs")); 1335 (void) zfs_error(hdl, EZFS_BADVERSION, msg); 1336 break; 1337 1338 case EDOM: 1339 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1340 "root pool can not have multiple vdevs" 1341 " or separate logs")); 1342 (void) zfs_error(hdl, EZFS_POOL_NOTSUP, msg); 1343 break; 1344 1345 case ENOTBLK: 1346 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1347 "cache device must be a disk or disk slice")); 1348 (void) zfs_error(hdl, EZFS_BADDEV, msg); 1349 break; 1350 1351 default: 1352 (void) zpool_standard_error(hdl, errno, msg); 1353 } 1354 1355 ret = -1; 1356 } else { 1357 ret = 0; 1358 } 1359 1360 zcmd_free_nvlists(&zc); 1361 1362 return (ret); 1363 } 1364 1365 /* 1366 * Exports the pool from the system. The caller must ensure that there are no 1367 * mounted datasets in the pool. 1368 */ 1369 static int 1370 zpool_export_common(zpool_handle_t *zhp, boolean_t force, boolean_t hardforce, 1371 const char *log_str) 1372 { 1373 zfs_cmd_t zc = { 0 }; 1374 char msg[1024]; 1375 1376 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 1377 "cannot export '%s'"), zhp->zpool_name); 1378 1379 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 1380 zc.zc_cookie = force; 1381 zc.zc_guid = hardforce; 1382 zc.zc_history = (uint64_t)(uintptr_t)log_str; 1383 1384 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_EXPORT, &zc) != 0) { 1385 switch (errno) { 1386 case EXDEV: 1387 zfs_error_aux(zhp->zpool_hdl, dgettext(TEXT_DOMAIN, 1388 "use '-f' to override the following errors:\n" 1389 "'%s' has an active shared spare which could be" 1390 " used by other pools once '%s' is exported."), 1391 zhp->zpool_name, zhp->zpool_name); 1392 return (zfs_error(zhp->zpool_hdl, EZFS_ACTIVE_SPARE, 1393 msg)); 1394 default: 1395 return (zpool_standard_error_fmt(zhp->zpool_hdl, errno, 1396 msg)); 1397 } 1398 } 1399 1400 return (0); 1401 } 1402 1403 int 1404 zpool_export(zpool_handle_t *zhp, boolean_t force, const char *log_str) 1405 { 1406 return (zpool_export_common(zhp, force, B_FALSE, log_str)); 1407 } 1408 1409 int 1410 zpool_export_force(zpool_handle_t *zhp, const char *log_str) 1411 { 1412 return (zpool_export_common(zhp, B_TRUE, B_TRUE, log_str)); 1413 } 1414 1415 static void 1416 zpool_rewind_exclaim(libzfs_handle_t *hdl, const char *name, boolean_t dryrun, 1417 nvlist_t *config) 1418 { 1419 nvlist_t *nv = NULL; 1420 uint64_t rewindto; 1421 int64_t loss = -1; 1422 struct tm t; 1423 char timestr[128]; 1424 1425 if (!hdl->libzfs_printerr || config == NULL) 1426 return; 1427 1428 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0 || 1429 nvlist_lookup_nvlist(nv, ZPOOL_CONFIG_REWIND_INFO, &nv) != 0) { 1430 return; 1431 } 1432 1433 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0) 1434 return; 1435 (void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss); 1436 1437 if (localtime_r((time_t *)&rewindto, &t) != NULL && 1438 strftime(timestr, 128, 0, &t) != 0) { 1439 if (dryrun) { 1440 (void) printf(dgettext(TEXT_DOMAIN, 1441 "Would be able to return %s " 1442 "to its state as of %s.\n"), 1443 name, timestr); 1444 } else { 1445 (void) printf(dgettext(TEXT_DOMAIN, 1446 "Pool %s returned to its state as of %s.\n"), 1447 name, timestr); 1448 } 1449 if (loss > 120) { 1450 (void) printf(dgettext(TEXT_DOMAIN, 1451 "%s approximately %lld "), 1452 dryrun ? "Would discard" : "Discarded", 1453 (loss + 30) / 60); 1454 (void) printf(dgettext(TEXT_DOMAIN, 1455 "minutes of transactions.\n")); 1456 } else if (loss > 0) { 1457 (void) printf(dgettext(TEXT_DOMAIN, 1458 "%s approximately %lld "), 1459 dryrun ? "Would discard" : "Discarded", loss); 1460 (void) printf(dgettext(TEXT_DOMAIN, 1461 "seconds of transactions.\n")); 1462 } 1463 } 1464 } 1465 1466 void 1467 zpool_explain_recover(libzfs_handle_t *hdl, const char *name, int reason, 1468 nvlist_t *config) 1469 { 1470 nvlist_t *nv = NULL; 1471 int64_t loss = -1; 1472 uint64_t edata = UINT64_MAX; 1473 uint64_t rewindto; 1474 struct tm t; 1475 char timestr[128]; 1476 1477 if (!hdl->libzfs_printerr) 1478 return; 1479 1480 if (reason >= 0) 1481 (void) printf(dgettext(TEXT_DOMAIN, "action: ")); 1482 else 1483 (void) printf(dgettext(TEXT_DOMAIN, "\t")); 1484 1485 /* All attempted rewinds failed if ZPOOL_CONFIG_LOAD_TIME missing */ 1486 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0 || 1487 nvlist_lookup_nvlist(nv, ZPOOL_CONFIG_REWIND_INFO, &nv) != 0 || 1488 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0) 1489 goto no_info; 1490 1491 (void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss); 1492 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_DATA_ERRORS, 1493 &edata); 1494 1495 (void) printf(dgettext(TEXT_DOMAIN, 1496 "Recovery is possible, but will result in some data loss.\n")); 1497 1498 if (localtime_r((time_t *)&rewindto, &t) != NULL && 1499 strftime(timestr, 128, 0, &t) != 0) { 1500 (void) printf(dgettext(TEXT_DOMAIN, 1501 "\tReturning the pool to its state as of %s\n" 1502 "\tshould correct the problem. "), 1503 timestr); 1504 } else { 1505 (void) printf(dgettext(TEXT_DOMAIN, 1506 "\tReverting the pool to an earlier state " 1507 "should correct the problem.\n\t")); 1508 } 1509 1510 if (loss > 120) { 1511 (void) printf(dgettext(TEXT_DOMAIN, 1512 "Approximately %lld minutes of data\n" 1513 "\tmust be discarded, irreversibly. "), (loss + 30) / 60); 1514 } else if (loss > 0) { 1515 (void) printf(dgettext(TEXT_DOMAIN, 1516 "Approximately %lld seconds of data\n" 1517 "\tmust be discarded, irreversibly. "), loss); 1518 } 1519 if (edata != 0 && edata != UINT64_MAX) { 1520 if (edata == 1) { 1521 (void) printf(dgettext(TEXT_DOMAIN, 1522 "After rewind, at least\n" 1523 "\tone persistent user-data error will remain. ")); 1524 } else { 1525 (void) printf(dgettext(TEXT_DOMAIN, 1526 "After rewind, several\n" 1527 "\tpersistent user-data errors will remain. ")); 1528 } 1529 } 1530 (void) printf(dgettext(TEXT_DOMAIN, 1531 "Recovery can be attempted\n\tby executing 'zpool %s -F %s'. "), 1532 reason >= 0 ? "clear" : "import", name); 1533 1534 (void) printf(dgettext(TEXT_DOMAIN, 1535 "A scrub of the pool\n" 1536 "\tis strongly recommended after recovery.\n")); 1537 return; 1538 1539 no_info: 1540 (void) printf(dgettext(TEXT_DOMAIN, 1541 "Destroy and re-create the pool from\n\ta backup source.\n")); 1542 } 1543 1544 /* 1545 * zpool_import() is a contracted interface. Should be kept the same 1546 * if possible. 1547 * 1548 * Applications should use zpool_import_props() to import a pool with 1549 * new properties value to be set. 1550 */ 1551 int 1552 zpool_import(libzfs_handle_t *hdl, nvlist_t *config, const char *newname, 1553 char *altroot) 1554 { 1555 nvlist_t *props = NULL; 1556 int ret; 1557 1558 if (altroot != NULL) { 1559 if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0) { 1560 return (zfs_error_fmt(hdl, EZFS_NOMEM, 1561 dgettext(TEXT_DOMAIN, "cannot import '%s'"), 1562 newname)); 1563 } 1564 1565 if (nvlist_add_string(props, 1566 zpool_prop_to_name(ZPOOL_PROP_ALTROOT), altroot) != 0 || 1567 nvlist_add_string(props, 1568 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE), "none") != 0) { 1569 nvlist_free(props); 1570 return (zfs_error_fmt(hdl, EZFS_NOMEM, 1571 dgettext(TEXT_DOMAIN, "cannot import '%s'"), 1572 newname)); 1573 } 1574 } 1575 1576 ret = zpool_import_props(hdl, config, newname, props, 1577 ZFS_IMPORT_NORMAL); 1578 if (props) 1579 nvlist_free(props); 1580 return (ret); 1581 } 1582 1583 static void 1584 print_vdev_tree(libzfs_handle_t *hdl, const char *name, nvlist_t *nv, 1585 int indent) 1586 { 1587 nvlist_t **child; 1588 uint_t c, children; 1589 char *vname; 1590 uint64_t is_log = 0; 1591 1592 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_LOG, 1593 &is_log); 1594 1595 if (name != NULL) 1596 (void) printf("\t%*s%s%s\n", indent, "", name, 1597 is_log ? " [log]" : ""); 1598 1599 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, 1600 &child, &children) != 0) 1601 return; 1602 1603 for (c = 0; c < children; c++) { 1604 vname = zpool_vdev_name(hdl, NULL, child[c], B_TRUE); 1605 print_vdev_tree(hdl, vname, child[c], indent + 2); 1606 free(vname); 1607 } 1608 } 1609 1610 void 1611 zpool_print_unsup_feat(nvlist_t *config) 1612 { 1613 nvlist_t *nvinfo, *unsup_feat; 1614 1615 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 1616 0); 1617 verify(nvlist_lookup_nvlist(nvinfo, ZPOOL_CONFIG_UNSUP_FEAT, 1618 &unsup_feat) == 0); 1619 1620 for (nvpair_t *nvp = nvlist_next_nvpair(unsup_feat, NULL); nvp != NULL; 1621 nvp = nvlist_next_nvpair(unsup_feat, nvp)) { 1622 char *desc; 1623 1624 verify(nvpair_type(nvp) == DATA_TYPE_STRING); 1625 verify(nvpair_value_string(nvp, &desc) == 0); 1626 1627 if (strlen(desc) > 0) 1628 (void) printf("\t%s (%s)\n", nvpair_name(nvp), desc); 1629 else 1630 (void) printf("\t%s\n", nvpair_name(nvp)); 1631 } 1632 } 1633 1634 /* 1635 * Import the given pool using the known configuration and a list of 1636 * properties to be set. The configuration should have come from 1637 * zpool_find_import(). The 'newname' parameters control whether the pool 1638 * is imported with a different name. 1639 */ 1640 int 1641 zpool_import_props(libzfs_handle_t *hdl, nvlist_t *config, const char *newname, 1642 nvlist_t *props, int flags) 1643 { 1644 zfs_cmd_t zc = { 0 }; 1645 zpool_rewind_policy_t policy; 1646 nvlist_t *nv = NULL; 1647 nvlist_t *nvinfo = NULL; 1648 nvlist_t *missing = NULL; 1649 char *thename; 1650 char *origname; 1651 int ret; 1652 int error = 0; 1653 char errbuf[1024]; 1654 1655 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME, 1656 &origname) == 0); 1657 1658 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, 1659 "cannot import pool '%s'"), origname); 1660 1661 if (newname != NULL) { 1662 if (!zpool_name_valid(hdl, B_FALSE, newname)) 1663 return (zfs_error_fmt(hdl, EZFS_INVALIDNAME, 1664 dgettext(TEXT_DOMAIN, "cannot import '%s'"), 1665 newname)); 1666 thename = (char *)newname; 1667 } else { 1668 thename = origname; 1669 } 1670 1671 if (props) { 1672 uint64_t version; 1673 prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE }; 1674 1675 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION, 1676 &version) == 0); 1677 1678 if ((props = zpool_valid_proplist(hdl, origname, 1679 props, version, flags, errbuf)) == NULL) { 1680 return (-1); 1681 } else if (zcmd_write_src_nvlist(hdl, &zc, props) != 0) { 1682 nvlist_free(props); 1683 return (-1); 1684 } 1685 } 1686 1687 (void) strlcpy(zc.zc_name, thename, sizeof (zc.zc_name)); 1688 1689 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, 1690 &zc.zc_guid) == 0); 1691 1692 if (zcmd_write_conf_nvlist(hdl, &zc, config) != 0) { 1693 nvlist_free(props); 1694 return (-1); 1695 } 1696 if (zcmd_alloc_dst_nvlist(hdl, &zc, zc.zc_nvlist_conf_size * 2) != 0) { 1697 nvlist_free(props); 1698 return (-1); 1699 } 1700 1701 zc.zc_cookie = flags; 1702 while ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_IMPORT, &zc)) != 0 && 1703 errno == ENOMEM) { 1704 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) { 1705 zcmd_free_nvlists(&zc); 1706 return (-1); 1707 } 1708 } 1709 if (ret != 0) 1710 error = errno; 1711 1712 (void) zcmd_read_dst_nvlist(hdl, &zc, &nv); 1713 zpool_get_rewind_policy(config, &policy); 1714 1715 if (error) { 1716 char desc[1024]; 1717 1718 /* 1719 * Dry-run failed, but we print out what success 1720 * looks like if we found a best txg 1721 */ 1722 if (policy.zrp_request & ZPOOL_TRY_REWIND) { 1723 zpool_rewind_exclaim(hdl, newname ? origname : thename, 1724 B_TRUE, nv); 1725 nvlist_free(nv); 1726 return (-1); 1727 } 1728 1729 if (newname == NULL) 1730 (void) snprintf(desc, sizeof (desc), 1731 dgettext(TEXT_DOMAIN, "cannot import '%s'"), 1732 thename); 1733 else 1734 (void) snprintf(desc, sizeof (desc), 1735 dgettext(TEXT_DOMAIN, "cannot import '%s' as '%s'"), 1736 origname, thename); 1737 1738 switch (error) { 1739 case ENOTSUP: 1740 if (nv != NULL && nvlist_lookup_nvlist(nv, 1741 ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0 && 1742 nvlist_exists(nvinfo, ZPOOL_CONFIG_UNSUP_FEAT)) { 1743 (void) printf(dgettext(TEXT_DOMAIN, "This " 1744 "pool uses the following feature(s) not " 1745 "supported by this system:\n")); 1746 zpool_print_unsup_feat(nv); 1747 if (nvlist_exists(nvinfo, 1748 ZPOOL_CONFIG_CAN_RDONLY)) { 1749 (void) printf(dgettext(TEXT_DOMAIN, 1750 "All unsupported features are only " 1751 "required for writing to the pool." 1752 "\nThe pool can be imported using " 1753 "'-o readonly=on'.\n")); 1754 } 1755 } 1756 /* 1757 * Unsupported version. 1758 */ 1759 (void) zfs_error(hdl, EZFS_BADVERSION, desc); 1760 break; 1761 1762 case EINVAL: 1763 (void) zfs_error(hdl, EZFS_INVALCONFIG, desc); 1764 break; 1765 1766 case EROFS: 1767 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1768 "one or more devices is read only")); 1769 (void) zfs_error(hdl, EZFS_BADDEV, desc); 1770 break; 1771 1772 case ENXIO: 1773 if (nv && nvlist_lookup_nvlist(nv, 1774 ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0 && 1775 nvlist_lookup_nvlist(nvinfo, 1776 ZPOOL_CONFIG_MISSING_DEVICES, &missing) == 0) { 1777 (void) printf(dgettext(TEXT_DOMAIN, 1778 "The devices below are missing, use " 1779 "'-m' to import the pool anyway:\n")); 1780 print_vdev_tree(hdl, NULL, missing, 2); 1781 (void) printf("\n"); 1782 } 1783 (void) zpool_standard_error(hdl, error, desc); 1784 break; 1785 1786 case EEXIST: 1787 (void) zpool_standard_error(hdl, error, desc); 1788 break; 1789 1790 default: 1791 (void) zpool_standard_error(hdl, error, desc); 1792 zpool_explain_recover(hdl, 1793 newname ? origname : thename, -error, nv); 1794 break; 1795 } 1796 1797 nvlist_free(nv); 1798 ret = -1; 1799 } else { 1800 zpool_handle_t *zhp; 1801 1802 /* 1803 * This should never fail, but play it safe anyway. 1804 */ 1805 if (zpool_open_silent(hdl, thename, &zhp) != 0) 1806 ret = -1; 1807 else if (zhp != NULL) 1808 zpool_close(zhp); 1809 if (policy.zrp_request & 1810 (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) { 1811 zpool_rewind_exclaim(hdl, newname ? origname : thename, 1812 ((policy.zrp_request & ZPOOL_TRY_REWIND) != 0), nv); 1813 } 1814 nvlist_free(nv); 1815 return (0); 1816 } 1817 1818 zcmd_free_nvlists(&zc); 1819 nvlist_free(props); 1820 1821 return (ret); 1822 } 1823 1824 /* 1825 * Scan the pool. 1826 */ 1827 int 1828 zpool_scan(zpool_handle_t *zhp, pool_scan_func_t func) 1829 { 1830 zfs_cmd_t zc = { 0 }; 1831 char msg[1024]; 1832 libzfs_handle_t *hdl = zhp->zpool_hdl; 1833 1834 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 1835 zc.zc_cookie = func; 1836 1837 if (zfs_ioctl(hdl, ZFS_IOC_POOL_SCAN, &zc) == 0 || 1838 (errno == ENOENT && func != POOL_SCAN_NONE)) 1839 return (0); 1840 1841 if (func == POOL_SCAN_SCRUB) { 1842 (void) snprintf(msg, sizeof (msg), 1843 dgettext(TEXT_DOMAIN, "cannot scrub %s"), zc.zc_name); 1844 } else if (func == POOL_SCAN_NONE) { 1845 (void) snprintf(msg, sizeof (msg), 1846 dgettext(TEXT_DOMAIN, "cannot cancel scrubbing %s"), 1847 zc.zc_name); 1848 } else { 1849 assert(!"unexpected result"); 1850 } 1851 1852 if (errno == EBUSY) { 1853 nvlist_t *nvroot; 1854 pool_scan_stat_t *ps = NULL; 1855 uint_t psc; 1856 1857 verify(nvlist_lookup_nvlist(zhp->zpool_config, 1858 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); 1859 (void) nvlist_lookup_uint64_array(nvroot, 1860 ZPOOL_CONFIG_SCAN_STATS, (uint64_t **)&ps, &psc); 1861 if (ps && ps->pss_func == POOL_SCAN_SCRUB) 1862 return (zfs_error(hdl, EZFS_SCRUBBING, msg)); 1863 else 1864 return (zfs_error(hdl, EZFS_RESILVERING, msg)); 1865 } else if (errno == ENOENT) { 1866 return (zfs_error(hdl, EZFS_NO_SCRUB, msg)); 1867 } else { 1868 return (zpool_standard_error(hdl, errno, msg)); 1869 } 1870 } 1871 1872 /* 1873 * This provides a very minimal check whether a given string is likely a 1874 * c#t#d# style string. Users of this are expected to do their own 1875 * verification of the s# part. 1876 */ 1877 #define CTD_CHECK(str) (str && str[0] == 'c' && isdigit(str[1])) 1878 1879 /* 1880 * More elaborate version for ones which may start with "/dev/dsk/" 1881 * and the like. 1882 */ 1883 static int 1884 ctd_check_path(char *str) { 1885 /* 1886 * If it starts with a slash, check the last component. 1887 */ 1888 if (str && str[0] == '/') { 1889 char *tmp = strrchr(str, '/'); 1890 1891 /* 1892 * If it ends in "/old", check the second-to-last 1893 * component of the string instead. 1894 */ 1895 if (tmp != str && strcmp(tmp, "/old") == 0) { 1896 for (tmp--; *tmp != '/'; tmp--) 1897 ; 1898 } 1899 str = tmp + 1; 1900 } 1901 return (CTD_CHECK(str)); 1902 } 1903 1904 /* 1905 * Find a vdev that matches the search criteria specified. We use the 1906 * the nvpair name to determine how we should look for the device. 1907 * 'avail_spare' is set to TRUE if the provided guid refers to an AVAIL 1908 * spare; but FALSE if its an INUSE spare. 1909 */ 1910 static nvlist_t * 1911 vdev_to_nvlist_iter(nvlist_t *nv, nvlist_t *search, boolean_t *avail_spare, 1912 boolean_t *l2cache, boolean_t *log) 1913 { 1914 uint_t c, children; 1915 nvlist_t **child; 1916 nvlist_t *ret; 1917 uint64_t is_log; 1918 char *srchkey; 1919 nvpair_t *pair = nvlist_next_nvpair(search, NULL); 1920 1921 /* Nothing to look for */ 1922 if (search == NULL || pair == NULL) 1923 return (NULL); 1924 1925 /* Obtain the key we will use to search */ 1926 srchkey = nvpair_name(pair); 1927 1928 switch (nvpair_type(pair)) { 1929 case DATA_TYPE_UINT64: 1930 if (strcmp(srchkey, ZPOOL_CONFIG_GUID) == 0) { 1931 uint64_t srchval, theguid; 1932 1933 verify(nvpair_value_uint64(pair, &srchval) == 0); 1934 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, 1935 &theguid) == 0); 1936 if (theguid == srchval) 1937 return (nv); 1938 } 1939 break; 1940 1941 case DATA_TYPE_STRING: { 1942 char *srchval, *val; 1943 1944 verify(nvpair_value_string(pair, &srchval) == 0); 1945 if (nvlist_lookup_string(nv, srchkey, &val) != 0) 1946 break; 1947 1948 /* 1949 * Search for the requested value. Special cases: 1950 * 1951 * - ZPOOL_CONFIG_PATH for whole disk entries. These end in 1952 * "s0" or "s0/old". The "s0" part is hidden from the user, 1953 * but included in the string, so this matches around it. 1954 * - looking for a top-level vdev name (i.e. ZPOOL_CONFIG_TYPE). 1955 * 1956 * Otherwise, all other searches are simple string compares. 1957 */ 1958 if (strcmp(srchkey, ZPOOL_CONFIG_PATH) == 0 && 1959 ctd_check_path(val)) { 1960 uint64_t wholedisk = 0; 1961 1962 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK, 1963 &wholedisk); 1964 if (wholedisk) { 1965 int slen = strlen(srchval); 1966 int vlen = strlen(val); 1967 1968 if (slen != vlen - 2) 1969 break; 1970 1971 /* 1972 * make_leaf_vdev() should only set 1973 * wholedisk for ZPOOL_CONFIG_PATHs which 1974 * will include "/dev/dsk/", giving plenty of 1975 * room for the indices used next. 1976 */ 1977 ASSERT(vlen >= 6); 1978 1979 /* 1980 * strings identical except trailing "s0" 1981 */ 1982 if (strcmp(&val[vlen - 2], "s0") == 0 && 1983 strncmp(srchval, val, slen) == 0) 1984 return (nv); 1985 1986 /* 1987 * strings identical except trailing "s0/old" 1988 */ 1989 if (strcmp(&val[vlen - 6], "s0/old") == 0 && 1990 strcmp(&srchval[slen - 4], "/old") == 0 && 1991 strncmp(srchval, val, slen - 4) == 0) 1992 return (nv); 1993 1994 break; 1995 } 1996 } else if (strcmp(srchkey, ZPOOL_CONFIG_TYPE) == 0 && val) { 1997 char *type, *idx, *end, *p; 1998 uint64_t id, vdev_id; 1999 2000 /* 2001 * Determine our vdev type, keeping in mind 2002 * that the srchval is composed of a type and 2003 * vdev id pair (i.e. mirror-4). 2004 */ 2005 if ((type = strdup(srchval)) == NULL) 2006 return (NULL); 2007 2008 if ((p = strrchr(type, '-')) == NULL) { 2009 free(type); 2010 break; 2011 } 2012 idx = p + 1; 2013 *p = '\0'; 2014 2015 /* 2016 * If the types don't match then keep looking. 2017 */ 2018 if (strncmp(val, type, strlen(val)) != 0) { 2019 free(type); 2020 break; 2021 } 2022 2023 verify(strncmp(type, VDEV_TYPE_RAIDZ, 2024 strlen(VDEV_TYPE_RAIDZ)) == 0 || 2025 strncmp(type, VDEV_TYPE_MIRROR, 2026 strlen(VDEV_TYPE_MIRROR)) == 0); 2027 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID, 2028 &id) == 0); 2029 2030 errno = 0; 2031 vdev_id = strtoull(idx, &end, 10); 2032 2033 free(type); 2034 if (errno != 0) 2035 return (NULL); 2036 2037 /* 2038 * Now verify that we have the correct vdev id. 2039 */ 2040 if (vdev_id == id) 2041 return (nv); 2042 } 2043 2044 /* 2045 * Common case 2046 */ 2047 if (strcmp(srchval, val) == 0) 2048 return (nv); 2049 break; 2050 } 2051 2052 default: 2053 break; 2054 } 2055 2056 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, 2057 &child, &children) != 0) 2058 return (NULL); 2059 2060 for (c = 0; c < children; c++) { 2061 if ((ret = vdev_to_nvlist_iter(child[c], search, 2062 avail_spare, l2cache, NULL)) != NULL) { 2063 /* 2064 * The 'is_log' value is only set for the toplevel 2065 * vdev, not the leaf vdevs. So we always lookup the 2066 * log device from the root of the vdev tree (where 2067 * 'log' is non-NULL). 2068 */ 2069 if (log != NULL && 2070 nvlist_lookup_uint64(child[c], 2071 ZPOOL_CONFIG_IS_LOG, &is_log) == 0 && 2072 is_log) { 2073 *log = B_TRUE; 2074 } 2075 return (ret); 2076 } 2077 } 2078 2079 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES, 2080 &child, &children) == 0) { 2081 for (c = 0; c < children; c++) { 2082 if ((ret = vdev_to_nvlist_iter(child[c], search, 2083 avail_spare, l2cache, NULL)) != NULL) { 2084 *avail_spare = B_TRUE; 2085 return (ret); 2086 } 2087 } 2088 } 2089 2090 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE, 2091 &child, &children) == 0) { 2092 for (c = 0; c < children; c++) { 2093 if ((ret = vdev_to_nvlist_iter(child[c], search, 2094 avail_spare, l2cache, NULL)) != NULL) { 2095 *l2cache = B_TRUE; 2096 return (ret); 2097 } 2098 } 2099 } 2100 2101 return (NULL); 2102 } 2103 2104 /* 2105 * Given a physical path (minus the "/devices" prefix), find the 2106 * associated vdev. 2107 */ 2108 nvlist_t * 2109 zpool_find_vdev_by_physpath(zpool_handle_t *zhp, const char *ppath, 2110 boolean_t *avail_spare, boolean_t *l2cache, boolean_t *log) 2111 { 2112 nvlist_t *search, *nvroot, *ret; 2113 2114 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2115 verify(nvlist_add_string(search, ZPOOL_CONFIG_PHYS_PATH, ppath) == 0); 2116 2117 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE, 2118 &nvroot) == 0); 2119 2120 *avail_spare = B_FALSE; 2121 *l2cache = B_FALSE; 2122 if (log != NULL) 2123 *log = B_FALSE; 2124 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log); 2125 nvlist_free(search); 2126 2127 return (ret); 2128 } 2129 2130 /* 2131 * Determine if we have an "interior" top-level vdev (i.e mirror/raidz). 2132 */ 2133 boolean_t 2134 zpool_vdev_is_interior(const char *name) 2135 { 2136 if (strncmp(name, VDEV_TYPE_RAIDZ, strlen(VDEV_TYPE_RAIDZ)) == 0 || 2137 strncmp(name, VDEV_TYPE_MIRROR, strlen(VDEV_TYPE_MIRROR)) == 0) 2138 return (B_TRUE); 2139 return (B_FALSE); 2140 } 2141 2142 nvlist_t * 2143 zpool_find_vdev(zpool_handle_t *zhp, const char *path, boolean_t *avail_spare, 2144 boolean_t *l2cache, boolean_t *log) 2145 { 2146 char buf[MAXPATHLEN]; 2147 char *end; 2148 nvlist_t *nvroot, *search, *ret; 2149 uint64_t guid; 2150 2151 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2152 2153 guid = strtoull(path, &end, 10); 2154 if (guid != 0 && *end == '\0') { 2155 verify(nvlist_add_uint64(search, ZPOOL_CONFIG_GUID, guid) == 0); 2156 } else if (zpool_vdev_is_interior(path)) { 2157 verify(nvlist_add_string(search, ZPOOL_CONFIG_TYPE, path) == 0); 2158 } else if (path[0] != '/') { 2159 (void) snprintf(buf, sizeof (buf), "%s%s", "/dev/dsk/", path); 2160 verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, buf) == 0); 2161 } else { 2162 verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, path) == 0); 2163 } 2164 2165 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE, 2166 &nvroot) == 0); 2167 2168 *avail_spare = B_FALSE; 2169 *l2cache = B_FALSE; 2170 if (log != NULL) 2171 *log = B_FALSE; 2172 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log); 2173 nvlist_free(search); 2174 2175 return (ret); 2176 } 2177 2178 static int 2179 vdev_online(nvlist_t *nv) 2180 { 2181 uint64_t ival; 2182 2183 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_OFFLINE, &ival) == 0 || 2184 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_FAULTED, &ival) == 0 || 2185 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_REMOVED, &ival) == 0) 2186 return (0); 2187 2188 return (1); 2189 } 2190 2191 /* 2192 * Helper function for zpool_get_physpaths(). 2193 */ 2194 static int 2195 vdev_get_one_physpath(nvlist_t *config, char *physpath, size_t physpath_size, 2196 size_t *bytes_written) 2197 { 2198 size_t bytes_left, pos, rsz; 2199 char *tmppath; 2200 const char *format; 2201 2202 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PHYS_PATH, 2203 &tmppath) != 0) 2204 return (EZFS_NODEVICE); 2205 2206 pos = *bytes_written; 2207 bytes_left = physpath_size - pos; 2208 format = (pos == 0) ? "%s" : " %s"; 2209 2210 rsz = snprintf(physpath + pos, bytes_left, format, tmppath); 2211 *bytes_written += rsz; 2212 2213 if (rsz >= bytes_left) { 2214 /* if physpath was not copied properly, clear it */ 2215 if (bytes_left != 0) { 2216 physpath[pos] = 0; 2217 } 2218 return (EZFS_NOSPC); 2219 } 2220 return (0); 2221 } 2222 2223 static int 2224 vdev_get_physpaths(nvlist_t *nv, char *physpath, size_t phypath_size, 2225 size_t *rsz, boolean_t is_spare) 2226 { 2227 char *type; 2228 int ret; 2229 2230 if (nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) != 0) 2231 return (EZFS_INVALCONFIG); 2232 2233 if (strcmp(type, VDEV_TYPE_DISK) == 0) { 2234 /* 2235 * An active spare device has ZPOOL_CONFIG_IS_SPARE set. 2236 * For a spare vdev, we only want to boot from the active 2237 * spare device. 2238 */ 2239 if (is_spare) { 2240 uint64_t spare = 0; 2241 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_SPARE, 2242 &spare); 2243 if (!spare) 2244 return (EZFS_INVALCONFIG); 2245 } 2246 2247 if (vdev_online(nv)) { 2248 if ((ret = vdev_get_one_physpath(nv, physpath, 2249 phypath_size, rsz)) != 0) 2250 return (ret); 2251 } 2252 } else if (strcmp(type, VDEV_TYPE_MIRROR) == 0 || 2253 strcmp(type, VDEV_TYPE_REPLACING) == 0 || 2254 (is_spare = (strcmp(type, VDEV_TYPE_SPARE) == 0))) { 2255 nvlist_t **child; 2256 uint_t count; 2257 int i, ret; 2258 2259 if (nvlist_lookup_nvlist_array(nv, 2260 ZPOOL_CONFIG_CHILDREN, &child, &count) != 0) 2261 return (EZFS_INVALCONFIG); 2262 2263 for (i = 0; i < count; i++) { 2264 ret = vdev_get_physpaths(child[i], physpath, 2265 phypath_size, rsz, is_spare); 2266 if (ret == EZFS_NOSPC) 2267 return (ret); 2268 } 2269 } 2270 2271 return (EZFS_POOL_INVALARG); 2272 } 2273 2274 /* 2275 * Get phys_path for a root pool config. 2276 * Return 0 on success; non-zero on failure. 2277 */ 2278 static int 2279 zpool_get_config_physpath(nvlist_t *config, char *physpath, size_t phypath_size) 2280 { 2281 size_t rsz; 2282 nvlist_t *vdev_root; 2283 nvlist_t **child; 2284 uint_t count; 2285 char *type; 2286 2287 rsz = 0; 2288 2289 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, 2290 &vdev_root) != 0) 2291 return (EZFS_INVALCONFIG); 2292 2293 if (nvlist_lookup_string(vdev_root, ZPOOL_CONFIG_TYPE, &type) != 0 || 2294 nvlist_lookup_nvlist_array(vdev_root, ZPOOL_CONFIG_CHILDREN, 2295 &child, &count) != 0) 2296 return (EZFS_INVALCONFIG); 2297 2298 /* 2299 * root pool can not have EFI labeled disks and can only have 2300 * a single top-level vdev. 2301 */ 2302 if (strcmp(type, VDEV_TYPE_ROOT) != 0 || count != 1 || 2303 pool_uses_efi(vdev_root)) 2304 return (EZFS_POOL_INVALARG); 2305 2306 (void) vdev_get_physpaths(child[0], physpath, phypath_size, &rsz, 2307 B_FALSE); 2308 2309 /* No online devices */ 2310 if (rsz == 0) 2311 return (EZFS_NODEVICE); 2312 2313 return (0); 2314 } 2315 2316 /* 2317 * Get phys_path for a root pool 2318 * Return 0 on success; non-zero on failure. 2319 */ 2320 int 2321 zpool_get_physpath(zpool_handle_t *zhp, char *physpath, size_t phypath_size) 2322 { 2323 return (zpool_get_config_physpath(zhp->zpool_config, physpath, 2324 phypath_size)); 2325 } 2326 2327 /* 2328 * If the device has being dynamically expanded then we need to relabel 2329 * the disk to use the new unallocated space. 2330 */ 2331 static int 2332 zpool_relabel_disk(libzfs_handle_t *hdl, const char *name) 2333 { 2334 char path[MAXPATHLEN]; 2335 char errbuf[1024]; 2336 int fd, error; 2337 int (*_efi_use_whole_disk)(int); 2338 2339 if ((_efi_use_whole_disk = (int (*)(int))dlsym(RTLD_DEFAULT, 2340 "efi_use_whole_disk")) == NULL) 2341 return (-1); 2342 2343 (void) snprintf(path, sizeof (path), "%s/%s", RDISK_ROOT, name); 2344 2345 if ((fd = open(path, O_RDWR | O_NDELAY)) < 0) { 2346 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot " 2347 "relabel '%s': unable to open device"), name); 2348 return (zfs_error(hdl, EZFS_OPENFAILED, errbuf)); 2349 } 2350 2351 /* 2352 * It's possible that we might encounter an error if the device 2353 * does not have any unallocated space left. If so, we simply 2354 * ignore that error and continue on. 2355 */ 2356 error = _efi_use_whole_disk(fd); 2357 (void) close(fd); 2358 if (error && error != VT_ENOSPC) { 2359 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot " 2360 "relabel '%s': unable to read disk capacity"), name); 2361 return (zfs_error(hdl, EZFS_NOCAP, errbuf)); 2362 } 2363 return (0); 2364 } 2365 2366 /* 2367 * Bring the specified vdev online. The 'flags' parameter is a set of the 2368 * ZFS_ONLINE_* flags. 2369 */ 2370 int 2371 zpool_vdev_online(zpool_handle_t *zhp, const char *path, int flags, 2372 vdev_state_t *newstate) 2373 { 2374 zfs_cmd_t zc = { 0 }; 2375 char msg[1024]; 2376 nvlist_t *tgt; 2377 boolean_t avail_spare, l2cache, islog; 2378 libzfs_handle_t *hdl = zhp->zpool_hdl; 2379 2380 if (flags & ZFS_ONLINE_EXPAND) { 2381 (void) snprintf(msg, sizeof (msg), 2382 dgettext(TEXT_DOMAIN, "cannot expand %s"), path); 2383 } else { 2384 (void) snprintf(msg, sizeof (msg), 2385 dgettext(TEXT_DOMAIN, "cannot online %s"), path); 2386 } 2387 2388 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2389 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache, 2390 &islog)) == NULL) 2391 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 2392 2393 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0); 2394 2395 if (avail_spare) 2396 return (zfs_error(hdl, EZFS_ISSPARE, msg)); 2397 2398 if (flags & ZFS_ONLINE_EXPAND || 2399 zpool_get_prop_int(zhp, ZPOOL_PROP_AUTOEXPAND, NULL)) { 2400 char *pathname = NULL; 2401 uint64_t wholedisk = 0; 2402 2403 (void) nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_WHOLE_DISK, 2404 &wholedisk); 2405 verify(nvlist_lookup_string(tgt, ZPOOL_CONFIG_PATH, 2406 &pathname) == 0); 2407 2408 /* 2409 * XXX - L2ARC 1.0 devices can't support expansion. 2410 */ 2411 if (l2cache) { 2412 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2413 "cannot expand cache devices")); 2414 return (zfs_error(hdl, EZFS_VDEVNOTSUP, msg)); 2415 } 2416 2417 if (wholedisk) { 2418 pathname += strlen(DISK_ROOT) + 1; 2419 (void) zpool_relabel_disk(hdl, pathname); 2420 } 2421 } 2422 2423 zc.zc_cookie = VDEV_STATE_ONLINE; 2424 zc.zc_obj = flags; 2425 2426 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) != 0) { 2427 if (errno == EINVAL) { 2428 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "was split " 2429 "from this pool into a new one. Use '%s' " 2430 "instead"), "zpool detach"); 2431 return (zfs_error(hdl, EZFS_POSTSPLIT_ONLINE, msg)); 2432 } 2433 return (zpool_standard_error(hdl, errno, msg)); 2434 } 2435 2436 *newstate = zc.zc_cookie; 2437 return (0); 2438 } 2439 2440 /* 2441 * Take the specified vdev offline 2442 */ 2443 int 2444 zpool_vdev_offline(zpool_handle_t *zhp, const char *path, boolean_t istmp) 2445 { 2446 zfs_cmd_t zc = { 0 }; 2447 char msg[1024]; 2448 nvlist_t *tgt; 2449 boolean_t avail_spare, l2cache; 2450 libzfs_handle_t *hdl = zhp->zpool_hdl; 2451 2452 (void) snprintf(msg, sizeof (msg), 2453 dgettext(TEXT_DOMAIN, "cannot offline %s"), path); 2454 2455 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2456 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache, 2457 NULL)) == NULL) 2458 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 2459 2460 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0); 2461 2462 if (avail_spare) 2463 return (zfs_error(hdl, EZFS_ISSPARE, msg)); 2464 2465 zc.zc_cookie = VDEV_STATE_OFFLINE; 2466 zc.zc_obj = istmp ? ZFS_OFFLINE_TEMPORARY : 0; 2467 2468 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) == 0) 2469 return (0); 2470 2471 switch (errno) { 2472 case EBUSY: 2473 2474 /* 2475 * There are no other replicas of this device. 2476 */ 2477 return (zfs_error(hdl, EZFS_NOREPLICAS, msg)); 2478 2479 case EEXIST: 2480 /* 2481 * The log device has unplayed logs 2482 */ 2483 return (zfs_error(hdl, EZFS_UNPLAYED_LOGS, msg)); 2484 2485 default: 2486 return (zpool_standard_error(hdl, errno, msg)); 2487 } 2488 } 2489 2490 /* 2491 * Mark the given vdev faulted. 2492 */ 2493 int 2494 zpool_vdev_fault(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux) 2495 { 2496 zfs_cmd_t zc = { 0 }; 2497 char msg[1024]; 2498 libzfs_handle_t *hdl = zhp->zpool_hdl; 2499 2500 (void) snprintf(msg, sizeof (msg), 2501 dgettext(TEXT_DOMAIN, "cannot fault %llu"), guid); 2502 2503 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2504 zc.zc_guid = guid; 2505 zc.zc_cookie = VDEV_STATE_FAULTED; 2506 zc.zc_obj = aux; 2507 2508 if (ioctl(hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0) 2509 return (0); 2510 2511 switch (errno) { 2512 case EBUSY: 2513 2514 /* 2515 * There are no other replicas of this device. 2516 */ 2517 return (zfs_error(hdl, EZFS_NOREPLICAS, msg)); 2518 2519 default: 2520 return (zpool_standard_error(hdl, errno, msg)); 2521 } 2522 2523 } 2524 2525 /* 2526 * Mark the given vdev degraded. 2527 */ 2528 int 2529 zpool_vdev_degrade(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux) 2530 { 2531 zfs_cmd_t zc = { 0 }; 2532 char msg[1024]; 2533 libzfs_handle_t *hdl = zhp->zpool_hdl; 2534 2535 (void) snprintf(msg, sizeof (msg), 2536 dgettext(TEXT_DOMAIN, "cannot degrade %llu"), guid); 2537 2538 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2539 zc.zc_guid = guid; 2540 zc.zc_cookie = VDEV_STATE_DEGRADED; 2541 zc.zc_obj = aux; 2542 2543 if (ioctl(hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0) 2544 return (0); 2545 2546 return (zpool_standard_error(hdl, errno, msg)); 2547 } 2548 2549 /* 2550 * Returns TRUE if the given nvlist is a vdev that was originally swapped in as 2551 * a hot spare. 2552 */ 2553 static boolean_t 2554 is_replacing_spare(nvlist_t *search, nvlist_t *tgt, int which) 2555 { 2556 nvlist_t **child; 2557 uint_t c, children; 2558 char *type; 2559 2560 if (nvlist_lookup_nvlist_array(search, ZPOOL_CONFIG_CHILDREN, &child, 2561 &children) == 0) { 2562 verify(nvlist_lookup_string(search, ZPOOL_CONFIG_TYPE, 2563 &type) == 0); 2564 2565 if (strcmp(type, VDEV_TYPE_SPARE) == 0 && 2566 children == 2 && child[which] == tgt) 2567 return (B_TRUE); 2568 2569 for (c = 0; c < children; c++) 2570 if (is_replacing_spare(child[c], tgt, which)) 2571 return (B_TRUE); 2572 } 2573 2574 return (B_FALSE); 2575 } 2576 2577 /* 2578 * Attach new_disk (fully described by nvroot) to old_disk. 2579 * If 'replacing' is specified, the new disk will replace the old one. 2580 */ 2581 int 2582 zpool_vdev_attach(zpool_handle_t *zhp, 2583 const char *old_disk, const char *new_disk, nvlist_t *nvroot, int replacing) 2584 { 2585 zfs_cmd_t zc = { 0 }; 2586 char msg[1024]; 2587 int ret; 2588 nvlist_t *tgt; 2589 boolean_t avail_spare, l2cache, islog; 2590 uint64_t val; 2591 char *newname; 2592 nvlist_t **child; 2593 uint_t children; 2594 nvlist_t *config_root; 2595 libzfs_handle_t *hdl = zhp->zpool_hdl; 2596 boolean_t rootpool = zpool_is_bootable(zhp); 2597 2598 if (replacing) 2599 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 2600 "cannot replace %s with %s"), old_disk, new_disk); 2601 else 2602 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 2603 "cannot attach %s to %s"), new_disk, old_disk); 2604 2605 /* 2606 * If this is a root pool, make sure that we're not attaching an 2607 * EFI labeled device. 2608 */ 2609 if (rootpool && pool_uses_efi(nvroot)) { 2610 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2611 "EFI labeled devices are not supported on root pools.")); 2612 return (zfs_error(hdl, EZFS_POOL_NOTSUP, msg)); 2613 } 2614 2615 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2616 if ((tgt = zpool_find_vdev(zhp, old_disk, &avail_spare, &l2cache, 2617 &islog)) == 0) 2618 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 2619 2620 if (avail_spare) 2621 return (zfs_error(hdl, EZFS_ISSPARE, msg)); 2622 2623 if (l2cache) 2624 return (zfs_error(hdl, EZFS_ISL2CACHE, msg)); 2625 2626 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0); 2627 zc.zc_cookie = replacing; 2628 2629 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, 2630 &child, &children) != 0 || children != 1) { 2631 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2632 "new device must be a single disk")); 2633 return (zfs_error(hdl, EZFS_INVALCONFIG, msg)); 2634 } 2635 2636 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL), 2637 ZPOOL_CONFIG_VDEV_TREE, &config_root) == 0); 2638 2639 if ((newname = zpool_vdev_name(NULL, NULL, child[0], B_FALSE)) == NULL) 2640 return (-1); 2641 2642 /* 2643 * If the target is a hot spare that has been swapped in, we can only 2644 * replace it with another hot spare. 2645 */ 2646 if (replacing && 2647 nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_IS_SPARE, &val) == 0 && 2648 (zpool_find_vdev(zhp, newname, &avail_spare, &l2cache, 2649 NULL) == NULL || !avail_spare) && 2650 is_replacing_spare(config_root, tgt, 1)) { 2651 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2652 "can only be replaced by another hot spare")); 2653 free(newname); 2654 return (zfs_error(hdl, EZFS_BADTARGET, msg)); 2655 } 2656 2657 free(newname); 2658 2659 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0) 2660 return (-1); 2661 2662 ret = zfs_ioctl(hdl, ZFS_IOC_VDEV_ATTACH, &zc); 2663 2664 zcmd_free_nvlists(&zc); 2665 2666 if (ret == 0) { 2667 if (rootpool) { 2668 /* 2669 * XXX need a better way to prevent user from 2670 * booting up a half-baked vdev. 2671 */ 2672 (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Make " 2673 "sure to wait until resilver is done " 2674 "before rebooting.\n")); 2675 } 2676 return (0); 2677 } 2678 2679 switch (errno) { 2680 case ENOTSUP: 2681 /* 2682 * Can't attach to or replace this type of vdev. 2683 */ 2684 if (replacing) { 2685 uint64_t version = zpool_get_prop_int(zhp, 2686 ZPOOL_PROP_VERSION, NULL); 2687 2688 if (islog) 2689 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2690 "cannot replace a log with a spare")); 2691 else if (version >= SPA_VERSION_MULTI_REPLACE) 2692 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2693 "already in replacing/spare config; wait " 2694 "for completion or use 'zpool detach'")); 2695 else 2696 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2697 "cannot replace a replacing device")); 2698 } else { 2699 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2700 "can only attach to mirrors and top-level " 2701 "disks")); 2702 } 2703 (void) zfs_error(hdl, EZFS_BADTARGET, msg); 2704 break; 2705 2706 case EINVAL: 2707 /* 2708 * The new device must be a single disk. 2709 */ 2710 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2711 "new device must be a single disk")); 2712 (void) zfs_error(hdl, EZFS_INVALCONFIG, msg); 2713 break; 2714 2715 case EBUSY: 2716 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "%s is busy"), 2717 new_disk); 2718 (void) zfs_error(hdl, EZFS_BADDEV, msg); 2719 break; 2720 2721 case EOVERFLOW: 2722 /* 2723 * The new device is too small. 2724 */ 2725 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2726 "device is too small")); 2727 (void) zfs_error(hdl, EZFS_BADDEV, msg); 2728 break; 2729 2730 case EDOM: 2731 /* 2732 * The new device has a different alignment requirement. 2733 */ 2734 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2735 "devices have different sector alignment")); 2736 (void) zfs_error(hdl, EZFS_BADDEV, msg); 2737 break; 2738 2739 case ENAMETOOLONG: 2740 /* 2741 * The resulting top-level vdev spec won't fit in the label. 2742 */ 2743 (void) zfs_error(hdl, EZFS_DEVOVERFLOW, msg); 2744 break; 2745 2746 default: 2747 (void) zpool_standard_error(hdl, errno, msg); 2748 } 2749 2750 return (-1); 2751 } 2752 2753 /* 2754 * Detach the specified device. 2755 */ 2756 int 2757 zpool_vdev_detach(zpool_handle_t *zhp, const char *path) 2758 { 2759 zfs_cmd_t zc = { 0 }; 2760 char msg[1024]; 2761 nvlist_t *tgt; 2762 boolean_t avail_spare, l2cache; 2763 libzfs_handle_t *hdl = zhp->zpool_hdl; 2764 2765 (void) snprintf(msg, sizeof (msg), 2766 dgettext(TEXT_DOMAIN, "cannot detach %s"), path); 2767 2768 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2769 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache, 2770 NULL)) == 0) 2771 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 2772 2773 if (avail_spare) 2774 return (zfs_error(hdl, EZFS_ISSPARE, msg)); 2775 2776 if (l2cache) 2777 return (zfs_error(hdl, EZFS_ISL2CACHE, msg)); 2778 2779 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0); 2780 2781 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_DETACH, &zc) == 0) 2782 return (0); 2783 2784 switch (errno) { 2785 2786 case ENOTSUP: 2787 /* 2788 * Can't detach from this type of vdev. 2789 */ 2790 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "only " 2791 "applicable to mirror and replacing vdevs")); 2792 (void) zfs_error(hdl, EZFS_BADTARGET, msg); 2793 break; 2794 2795 case EBUSY: 2796 /* 2797 * There are no other replicas of this device. 2798 */ 2799 (void) zfs_error(hdl, EZFS_NOREPLICAS, msg); 2800 break; 2801 2802 default: 2803 (void) zpool_standard_error(hdl, errno, msg); 2804 } 2805 2806 return (-1); 2807 } 2808 2809 /* 2810 * Find a mirror vdev in the source nvlist. 2811 * 2812 * The mchild array contains a list of disks in one of the top-level mirrors 2813 * of the source pool. The schild array contains a list of disks that the 2814 * user specified on the command line. We loop over the mchild array to 2815 * see if any entry in the schild array matches. 2816 * 2817 * If a disk in the mchild array is found in the schild array, we return 2818 * the index of that entry. Otherwise we return -1. 2819 */ 2820 static int 2821 find_vdev_entry(zpool_handle_t *zhp, nvlist_t **mchild, uint_t mchildren, 2822 nvlist_t **schild, uint_t schildren) 2823 { 2824 uint_t mc; 2825 2826 for (mc = 0; mc < mchildren; mc++) { 2827 uint_t sc; 2828 char *mpath = zpool_vdev_name(zhp->zpool_hdl, zhp, 2829 mchild[mc], B_FALSE); 2830 2831 for (sc = 0; sc < schildren; sc++) { 2832 char *spath = zpool_vdev_name(zhp->zpool_hdl, zhp, 2833 schild[sc], B_FALSE); 2834 boolean_t result = (strcmp(mpath, spath) == 0); 2835 2836 free(spath); 2837 if (result) { 2838 free(mpath); 2839 return (mc); 2840 } 2841 } 2842 2843 free(mpath); 2844 } 2845 2846 return (-1); 2847 } 2848 2849 /* 2850 * Split a mirror pool. If newroot points to null, then a new nvlist 2851 * is generated and it is the responsibility of the caller to free it. 2852 */ 2853 int 2854 zpool_vdev_split(zpool_handle_t *zhp, char *newname, nvlist_t **newroot, 2855 nvlist_t *props, splitflags_t flags) 2856 { 2857 zfs_cmd_t zc = { 0 }; 2858 char msg[1024]; 2859 nvlist_t *tree, *config, **child, **newchild, *newconfig = NULL; 2860 nvlist_t **varray = NULL, *zc_props = NULL; 2861 uint_t c, children, newchildren, lastlog = 0, vcount, found = 0; 2862 libzfs_handle_t *hdl = zhp->zpool_hdl; 2863 uint64_t vers; 2864 boolean_t freelist = B_FALSE, memory_err = B_TRUE; 2865 int retval = 0; 2866 2867 (void) snprintf(msg, sizeof (msg), 2868 dgettext(TEXT_DOMAIN, "Unable to split %s"), zhp->zpool_name); 2869 2870 if (!zpool_name_valid(hdl, B_FALSE, newname)) 2871 return (zfs_error(hdl, EZFS_INVALIDNAME, msg)); 2872 2873 if ((config = zpool_get_config(zhp, NULL)) == NULL) { 2874 (void) fprintf(stderr, gettext("Internal error: unable to " 2875 "retrieve pool configuration\n")); 2876 return (-1); 2877 } 2878 2879 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &tree) 2880 == 0); 2881 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION, &vers) == 0); 2882 2883 if (props) { 2884 prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE }; 2885 if ((zc_props = zpool_valid_proplist(hdl, zhp->zpool_name, 2886 props, vers, flags, msg)) == NULL) 2887 return (-1); 2888 } 2889 2890 if (nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN, &child, 2891 &children) != 0) { 2892 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2893 "Source pool is missing vdev tree")); 2894 if (zc_props) 2895 nvlist_free(zc_props); 2896 return (-1); 2897 } 2898 2899 varray = zfs_alloc(hdl, children * sizeof (nvlist_t *)); 2900 vcount = 0; 2901 2902 if (*newroot == NULL || 2903 nvlist_lookup_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN, 2904 &newchild, &newchildren) != 0) 2905 newchildren = 0; 2906 2907 for (c = 0; c < children; c++) { 2908 uint64_t is_log = B_FALSE, is_hole = B_FALSE; 2909 char *type; 2910 nvlist_t **mchild, *vdev; 2911 uint_t mchildren; 2912 int entry; 2913 2914 /* 2915 * Unlike cache & spares, slogs are stored in the 2916 * ZPOOL_CONFIG_CHILDREN array. We filter them out here. 2917 */ 2918 (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG, 2919 &is_log); 2920 (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE, 2921 &is_hole); 2922 if (is_log || is_hole) { 2923 /* 2924 * Create a hole vdev and put it in the config. 2925 */ 2926 if (nvlist_alloc(&vdev, NV_UNIQUE_NAME, 0) != 0) 2927 goto out; 2928 if (nvlist_add_string(vdev, ZPOOL_CONFIG_TYPE, 2929 VDEV_TYPE_HOLE) != 0) 2930 goto out; 2931 if (nvlist_add_uint64(vdev, ZPOOL_CONFIG_IS_HOLE, 2932 1) != 0) 2933 goto out; 2934 if (lastlog == 0) 2935 lastlog = vcount; 2936 varray[vcount++] = vdev; 2937 continue; 2938 } 2939 lastlog = 0; 2940 verify(nvlist_lookup_string(child[c], ZPOOL_CONFIG_TYPE, &type) 2941 == 0); 2942 if (strcmp(type, VDEV_TYPE_MIRROR) != 0) { 2943 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2944 "Source pool must be composed only of mirrors\n")); 2945 retval = zfs_error(hdl, EZFS_INVALCONFIG, msg); 2946 goto out; 2947 } 2948 2949 verify(nvlist_lookup_nvlist_array(child[c], 2950 ZPOOL_CONFIG_CHILDREN, &mchild, &mchildren) == 0); 2951 2952 /* find or add an entry for this top-level vdev */ 2953 if (newchildren > 0 && 2954 (entry = find_vdev_entry(zhp, mchild, mchildren, 2955 newchild, newchildren)) >= 0) { 2956 /* We found a disk that the user specified. */ 2957 vdev = mchild[entry]; 2958 ++found; 2959 } else { 2960 /* User didn't specify a disk for this vdev. */ 2961 vdev = mchild[mchildren - 1]; 2962 } 2963 2964 if (nvlist_dup(vdev, &varray[vcount++], 0) != 0) 2965 goto out; 2966 } 2967 2968 /* did we find every disk the user specified? */ 2969 if (found != newchildren) { 2970 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "Device list must " 2971 "include at most one disk from each mirror")); 2972 retval = zfs_error(hdl, EZFS_INVALCONFIG, msg); 2973 goto out; 2974 } 2975 2976 /* Prepare the nvlist for populating. */ 2977 if (*newroot == NULL) { 2978 if (nvlist_alloc(newroot, NV_UNIQUE_NAME, 0) != 0) 2979 goto out; 2980 freelist = B_TRUE; 2981 if (nvlist_add_string(*newroot, ZPOOL_CONFIG_TYPE, 2982 VDEV_TYPE_ROOT) != 0) 2983 goto out; 2984 } else { 2985 verify(nvlist_remove_all(*newroot, ZPOOL_CONFIG_CHILDREN) == 0); 2986 } 2987 2988 /* Add all the children we found */ 2989 if (nvlist_add_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN, varray, 2990 lastlog == 0 ? vcount : lastlog) != 0) 2991 goto out; 2992 2993 /* 2994 * If we're just doing a dry run, exit now with success. 2995 */ 2996 if (flags.dryrun) { 2997 memory_err = B_FALSE; 2998 freelist = B_FALSE; 2999 goto out; 3000 } 3001 3002 /* now build up the config list & call the ioctl */ 3003 if (nvlist_alloc(&newconfig, NV_UNIQUE_NAME, 0) != 0) 3004 goto out; 3005 3006 if (nvlist_add_nvlist(newconfig, 3007 ZPOOL_CONFIG_VDEV_TREE, *newroot) != 0 || 3008 nvlist_add_string(newconfig, 3009 ZPOOL_CONFIG_POOL_NAME, newname) != 0 || 3010 nvlist_add_uint64(newconfig, ZPOOL_CONFIG_VERSION, vers) != 0) 3011 goto out; 3012 3013 /* 3014 * The new pool is automatically part of the namespace unless we 3015 * explicitly export it. 3016 */ 3017 if (!flags.import) 3018 zc.zc_cookie = ZPOOL_EXPORT_AFTER_SPLIT; 3019 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3020 (void) strlcpy(zc.zc_string, newname, sizeof (zc.zc_string)); 3021 if (zcmd_write_conf_nvlist(hdl, &zc, newconfig) != 0) 3022 goto out; 3023 if (zc_props != NULL && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0) 3024 goto out; 3025 3026 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SPLIT, &zc) != 0) { 3027 retval = zpool_standard_error(hdl, errno, msg); 3028 goto out; 3029 } 3030 3031 freelist = B_FALSE; 3032 memory_err = B_FALSE; 3033 3034 out: 3035 if (varray != NULL) { 3036 int v; 3037 3038 for (v = 0; v < vcount; v++) 3039 nvlist_free(varray[v]); 3040 free(varray); 3041 } 3042 zcmd_free_nvlists(&zc); 3043 if (zc_props) 3044 nvlist_free(zc_props); 3045 if (newconfig) 3046 nvlist_free(newconfig); 3047 if (freelist) { 3048 nvlist_free(*newroot); 3049 *newroot = NULL; 3050 } 3051 3052 if (retval != 0) 3053 return (retval); 3054 3055 if (memory_err) 3056 return (no_memory(hdl)); 3057 3058 return (0); 3059 } 3060 3061 /* 3062 * Remove the given device. Currently, this is supported only for hot spares 3063 * and level 2 cache devices. 3064 */ 3065 int 3066 zpool_vdev_remove(zpool_handle_t *zhp, const char *path) 3067 { 3068 zfs_cmd_t zc = { 0 }; 3069 char msg[1024]; 3070 nvlist_t *tgt; 3071 boolean_t avail_spare, l2cache, islog; 3072 libzfs_handle_t *hdl = zhp->zpool_hdl; 3073 uint64_t version; 3074 3075 (void) snprintf(msg, sizeof (msg), 3076 dgettext(TEXT_DOMAIN, "cannot remove %s"), path); 3077 3078 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3079 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache, 3080 &islog)) == 0) 3081 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 3082 /* 3083 * XXX - this should just go away. 3084 */ 3085 if (!avail_spare && !l2cache && !islog) { 3086 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3087 "only inactive hot spares, cache, top-level, " 3088 "or log devices can be removed")); 3089 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 3090 } 3091 3092 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL); 3093 if (islog && version < SPA_VERSION_HOLES) { 3094 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3095 "pool must be upgrade to support log removal")); 3096 return (zfs_error(hdl, EZFS_BADVERSION, msg)); 3097 } 3098 3099 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0); 3100 3101 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_REMOVE, &zc) == 0) 3102 return (0); 3103 3104 return (zpool_standard_error(hdl, errno, msg)); 3105 } 3106 3107 /* 3108 * Clear the errors for the pool, or the particular device if specified. 3109 */ 3110 int 3111 zpool_clear(zpool_handle_t *zhp, const char *path, nvlist_t *rewindnvl) 3112 { 3113 zfs_cmd_t zc = { 0 }; 3114 char msg[1024]; 3115 nvlist_t *tgt; 3116 zpool_rewind_policy_t policy; 3117 boolean_t avail_spare, l2cache; 3118 libzfs_handle_t *hdl = zhp->zpool_hdl; 3119 nvlist_t *nvi = NULL; 3120 int error; 3121 3122 if (path) 3123 (void) snprintf(msg, sizeof (msg), 3124 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"), 3125 path); 3126 else 3127 (void) snprintf(msg, sizeof (msg), 3128 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"), 3129 zhp->zpool_name); 3130 3131 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3132 if (path) { 3133 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, 3134 &l2cache, NULL)) == 0) 3135 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 3136 3137 /* 3138 * Don't allow error clearing for hot spares. Do allow 3139 * error clearing for l2cache devices. 3140 */ 3141 if (avail_spare) 3142 return (zfs_error(hdl, EZFS_ISSPARE, msg)); 3143 3144 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, 3145 &zc.zc_guid) == 0); 3146 } 3147 3148 zpool_get_rewind_policy(rewindnvl, &policy); 3149 zc.zc_cookie = policy.zrp_request; 3150 3151 if (zcmd_alloc_dst_nvlist(hdl, &zc, zhp->zpool_config_size * 2) != 0) 3152 return (-1); 3153 3154 if (zcmd_write_src_nvlist(hdl, &zc, rewindnvl) != 0) 3155 return (-1); 3156 3157 while ((error = zfs_ioctl(hdl, ZFS_IOC_CLEAR, &zc)) != 0 && 3158 errno == ENOMEM) { 3159 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) { 3160 zcmd_free_nvlists(&zc); 3161 return (-1); 3162 } 3163 } 3164 3165 if (!error || ((policy.zrp_request & ZPOOL_TRY_REWIND) && 3166 errno != EPERM && errno != EACCES)) { 3167 if (policy.zrp_request & 3168 (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) { 3169 (void) zcmd_read_dst_nvlist(hdl, &zc, &nvi); 3170 zpool_rewind_exclaim(hdl, zc.zc_name, 3171 ((policy.zrp_request & ZPOOL_TRY_REWIND) != 0), 3172 nvi); 3173 nvlist_free(nvi); 3174 } 3175 zcmd_free_nvlists(&zc); 3176 return (0); 3177 } 3178 3179 zcmd_free_nvlists(&zc); 3180 return (zpool_standard_error(hdl, errno, msg)); 3181 } 3182 3183 /* 3184 * Similar to zpool_clear(), but takes a GUID (used by fmd). 3185 */ 3186 int 3187 zpool_vdev_clear(zpool_handle_t *zhp, uint64_t guid) 3188 { 3189 zfs_cmd_t zc = { 0 }; 3190 char msg[1024]; 3191 libzfs_handle_t *hdl = zhp->zpool_hdl; 3192 3193 (void) snprintf(msg, sizeof (msg), 3194 dgettext(TEXT_DOMAIN, "cannot clear errors for %llx"), 3195 guid); 3196 3197 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3198 zc.zc_guid = guid; 3199 zc.zc_cookie = ZPOOL_NO_REWIND; 3200 3201 if (ioctl(hdl->libzfs_fd, ZFS_IOC_CLEAR, &zc) == 0) 3202 return (0); 3203 3204 return (zpool_standard_error(hdl, errno, msg)); 3205 } 3206 3207 /* 3208 * Change the GUID for a pool. 3209 */ 3210 int 3211 zpool_reguid(zpool_handle_t *zhp) 3212 { 3213 char msg[1024]; 3214 libzfs_handle_t *hdl = zhp->zpool_hdl; 3215 zfs_cmd_t zc = { 0 }; 3216 3217 (void) snprintf(msg, sizeof (msg), 3218 dgettext(TEXT_DOMAIN, "cannot reguid '%s'"), zhp->zpool_name); 3219 3220 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3221 if (zfs_ioctl(hdl, ZFS_IOC_POOL_REGUID, &zc) == 0) 3222 return (0); 3223 3224 return (zpool_standard_error(hdl, errno, msg)); 3225 } 3226 3227 /* 3228 * Reopen the pool. 3229 */ 3230 int 3231 zpool_reopen(zpool_handle_t *zhp) 3232 { 3233 zfs_cmd_t zc = { 0 }; 3234 char msg[1024]; 3235 libzfs_handle_t *hdl = zhp->zpool_hdl; 3236 3237 (void) snprintf(msg, sizeof (msg), 3238 dgettext(TEXT_DOMAIN, "cannot reopen '%s'"), 3239 zhp->zpool_name); 3240 3241 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3242 if (zfs_ioctl(hdl, ZFS_IOC_POOL_REOPEN, &zc) == 0) 3243 return (0); 3244 return (zpool_standard_error(hdl, errno, msg)); 3245 } 3246 3247 /* 3248 * Convert from a devid string to a path. 3249 */ 3250 static char * 3251 devid_to_path(char *devid_str) 3252 { 3253 ddi_devid_t devid; 3254 char *minor; 3255 char *path; 3256 devid_nmlist_t *list = NULL; 3257 int ret; 3258 3259 if (devid_str_decode(devid_str, &devid, &minor) != 0) 3260 return (NULL); 3261 3262 ret = devid_deviceid_to_nmlist("/dev", devid, minor, &list); 3263 3264 devid_str_free(minor); 3265 devid_free(devid); 3266 3267 if (ret != 0) 3268 return (NULL); 3269 3270 if ((path = strdup(list[0].devname)) == NULL) 3271 return (NULL); 3272 3273 devid_free_nmlist(list); 3274 3275 return (path); 3276 } 3277 3278 /* 3279 * Convert from a path to a devid string. 3280 */ 3281 static char * 3282 path_to_devid(const char *path) 3283 { 3284 int fd; 3285 ddi_devid_t devid; 3286 char *minor, *ret; 3287 3288 if ((fd = open(path, O_RDONLY)) < 0) 3289 return (NULL); 3290 3291 minor = NULL; 3292 ret = NULL; 3293 if (devid_get(fd, &devid) == 0) { 3294 if (devid_get_minor_name(fd, &minor) == 0) 3295 ret = devid_str_encode(devid, minor); 3296 if (minor != NULL) 3297 devid_str_free(minor); 3298 devid_free(devid); 3299 } 3300 (void) close(fd); 3301 3302 return (ret); 3303 } 3304 3305 /* 3306 * Issue the necessary ioctl() to update the stored path value for the vdev. We 3307 * ignore any failure here, since a common case is for an unprivileged user to 3308 * type 'zpool status', and we'll display the correct information anyway. 3309 */ 3310 static void 3311 set_path(zpool_handle_t *zhp, nvlist_t *nv, const char *path) 3312 { 3313 zfs_cmd_t zc = { 0 }; 3314 3315 (void) strncpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3316 (void) strncpy(zc.zc_value, path, sizeof (zc.zc_value)); 3317 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, 3318 &zc.zc_guid) == 0); 3319 3320 (void) ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_VDEV_SETPATH, &zc); 3321 } 3322 3323 /* 3324 * Given a vdev, return the name to display in iostat. If the vdev has a path, 3325 * we use that, stripping off any leading "/dev/dsk/"; if not, we use the type. 3326 * We also check if this is a whole disk, in which case we strip off the 3327 * trailing 's0' slice name. 3328 * 3329 * This routine is also responsible for identifying when disks have been 3330 * reconfigured in a new location. The kernel will have opened the device by 3331 * devid, but the path will still refer to the old location. To catch this, we 3332 * first do a path -> devid translation (which is fast for the common case). If 3333 * the devid matches, we're done. If not, we do a reverse devid -> path 3334 * translation and issue the appropriate ioctl() to update the path of the vdev. 3335 * If 'zhp' is NULL, then this is an exported pool, and we don't need to do any 3336 * of these checks. 3337 */ 3338 char * 3339 zpool_vdev_name(libzfs_handle_t *hdl, zpool_handle_t *zhp, nvlist_t *nv, 3340 boolean_t verbose) 3341 { 3342 char *path, *devid; 3343 uint64_t value; 3344 char buf[64]; 3345 vdev_stat_t *vs; 3346 uint_t vsc; 3347 3348 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT, 3349 &value) == 0) { 3350 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, 3351 &value) == 0); 3352 (void) snprintf(buf, sizeof (buf), "%llu", 3353 (u_longlong_t)value); 3354 path = buf; 3355 } else if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0) { 3356 3357 /* 3358 * If the device is dead (faulted, offline, etc) then don't 3359 * bother opening it. Otherwise we may be forcing the user to 3360 * open a misbehaving device, which can have undesirable 3361 * effects. 3362 */ 3363 if ((nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS, 3364 (uint64_t **)&vs, &vsc) != 0 || 3365 vs->vs_state >= VDEV_STATE_DEGRADED) && 3366 zhp != NULL && 3367 nvlist_lookup_string(nv, ZPOOL_CONFIG_DEVID, &devid) == 0) { 3368 /* 3369 * Determine if the current path is correct. 3370 */ 3371 char *newdevid = path_to_devid(path); 3372 3373 if (newdevid == NULL || 3374 strcmp(devid, newdevid) != 0) { 3375 char *newpath; 3376 3377 if ((newpath = devid_to_path(devid)) != NULL) { 3378 /* 3379 * Update the path appropriately. 3380 */ 3381 set_path(zhp, nv, newpath); 3382 if (nvlist_add_string(nv, 3383 ZPOOL_CONFIG_PATH, newpath) == 0) 3384 verify(nvlist_lookup_string(nv, 3385 ZPOOL_CONFIG_PATH, 3386 &path) == 0); 3387 free(newpath); 3388 } 3389 } 3390 3391 if (newdevid) 3392 devid_str_free(newdevid); 3393 } 3394 3395 if (strncmp(path, "/dev/dsk/", 9) == 0) 3396 path += 9; 3397 3398 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK, 3399 &value) == 0 && value) { 3400 int pathlen = strlen(path); 3401 char *tmp = zfs_strdup(hdl, path); 3402 3403 /* 3404 * If it starts with c#, and ends with "s0", chop 3405 * the "s0" off, or if it ends with "s0/old", remove 3406 * the "s0" from the middle. 3407 */ 3408 if (CTD_CHECK(tmp)) { 3409 if (strcmp(&tmp[pathlen - 2], "s0") == 0) { 3410 tmp[pathlen - 2] = '\0'; 3411 } else if (pathlen > 6 && 3412 strcmp(&tmp[pathlen - 6], "s0/old") == 0) { 3413 (void) strcpy(&tmp[pathlen - 6], 3414 "/old"); 3415 } 3416 } 3417 return (tmp); 3418 } 3419 } else { 3420 verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &path) == 0); 3421 3422 /* 3423 * If it's a raidz device, we need to stick in the parity level. 3424 */ 3425 if (strcmp(path, VDEV_TYPE_RAIDZ) == 0) { 3426 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NPARITY, 3427 &value) == 0); 3428 (void) snprintf(buf, sizeof (buf), "%s%llu", path, 3429 (u_longlong_t)value); 3430 path = buf; 3431 } 3432 3433 /* 3434 * We identify each top-level vdev by using a <type-id> 3435 * naming convention. 3436 */ 3437 if (verbose) { 3438 uint64_t id; 3439 3440 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID, 3441 &id) == 0); 3442 (void) snprintf(buf, sizeof (buf), "%s-%llu", path, 3443 (u_longlong_t)id); 3444 path = buf; 3445 } 3446 } 3447 3448 return (zfs_strdup(hdl, path)); 3449 } 3450 3451 static int 3452 zbookmark_compare(const void *a, const void *b) 3453 { 3454 return (memcmp(a, b, sizeof (zbookmark_phys_t))); 3455 } 3456 3457 /* 3458 * Retrieve the persistent error log, uniquify the members, and return to the 3459 * caller. 3460 */ 3461 int 3462 zpool_get_errlog(zpool_handle_t *zhp, nvlist_t **nverrlistp) 3463 { 3464 zfs_cmd_t zc = { 0 }; 3465 uint64_t count; 3466 zbookmark_phys_t *zb = NULL; 3467 int i; 3468 3469 /* 3470 * Retrieve the raw error list from the kernel. If the number of errors 3471 * has increased, allocate more space and continue until we get the 3472 * entire list. 3473 */ 3474 verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_ERRCOUNT, 3475 &count) == 0); 3476 if (count == 0) 3477 return (0); 3478 if ((zc.zc_nvlist_dst = (uintptr_t)zfs_alloc(zhp->zpool_hdl, 3479 count * sizeof (zbookmark_phys_t))) == (uintptr_t)NULL) 3480 return (-1); 3481 zc.zc_nvlist_dst_size = count; 3482 (void) strcpy(zc.zc_name, zhp->zpool_name); 3483 for (;;) { 3484 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_ERROR_LOG, 3485 &zc) != 0) { 3486 free((void *)(uintptr_t)zc.zc_nvlist_dst); 3487 if (errno == ENOMEM) { 3488 void *dst; 3489 3490 count = zc.zc_nvlist_dst_size; 3491 dst = zfs_alloc(zhp->zpool_hdl, count * 3492 sizeof (zbookmark_phys_t)); 3493 if (dst == NULL) 3494 return (-1); 3495 zc.zc_nvlist_dst = (uintptr_t)dst; 3496 } else { 3497 return (-1); 3498 } 3499 } else { 3500 break; 3501 } 3502 } 3503 3504 /* 3505 * Sort the resulting bookmarks. This is a little confusing due to the 3506 * implementation of ZFS_IOC_ERROR_LOG. The bookmarks are copied last 3507 * to first, and 'zc_nvlist_dst_size' indicates the number of boomarks 3508 * _not_ copied as part of the process. So we point the start of our 3509 * array appropriate and decrement the total number of elements. 3510 */ 3511 zb = ((zbookmark_phys_t *)(uintptr_t)zc.zc_nvlist_dst) + 3512 zc.zc_nvlist_dst_size; 3513 count -= zc.zc_nvlist_dst_size; 3514 3515 qsort(zb, count, sizeof (zbookmark_phys_t), zbookmark_compare); 3516 3517 verify(nvlist_alloc(nverrlistp, 0, KM_SLEEP) == 0); 3518 3519 /* 3520 * Fill in the nverrlistp with nvlist's of dataset and object numbers. 3521 */ 3522 for (i = 0; i < count; i++) { 3523 nvlist_t *nv; 3524 3525 /* ignoring zb_blkid and zb_level for now */ 3526 if (i > 0 && zb[i-1].zb_objset == zb[i].zb_objset && 3527 zb[i-1].zb_object == zb[i].zb_object) 3528 continue; 3529 3530 if (nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) != 0) 3531 goto nomem; 3532 if (nvlist_add_uint64(nv, ZPOOL_ERR_DATASET, 3533 zb[i].zb_objset) != 0) { 3534 nvlist_free(nv); 3535 goto nomem; 3536 } 3537 if (nvlist_add_uint64(nv, ZPOOL_ERR_OBJECT, 3538 zb[i].zb_object) != 0) { 3539 nvlist_free(nv); 3540 goto nomem; 3541 } 3542 if (nvlist_add_nvlist(*nverrlistp, "ejk", nv) != 0) { 3543 nvlist_free(nv); 3544 goto nomem; 3545 } 3546 nvlist_free(nv); 3547 } 3548 3549 free((void *)(uintptr_t)zc.zc_nvlist_dst); 3550 return (0); 3551 3552 nomem: 3553 free((void *)(uintptr_t)zc.zc_nvlist_dst); 3554 return (no_memory(zhp->zpool_hdl)); 3555 } 3556 3557 /* 3558 * Upgrade a ZFS pool to the latest on-disk version. 3559 */ 3560 int 3561 zpool_upgrade(zpool_handle_t *zhp, uint64_t new_version) 3562 { 3563 zfs_cmd_t zc = { 0 }; 3564 libzfs_handle_t *hdl = zhp->zpool_hdl; 3565 3566 (void) strcpy(zc.zc_name, zhp->zpool_name); 3567 zc.zc_cookie = new_version; 3568 3569 if (zfs_ioctl(hdl, ZFS_IOC_POOL_UPGRADE, &zc) != 0) 3570 return (zpool_standard_error_fmt(hdl, errno, 3571 dgettext(TEXT_DOMAIN, "cannot upgrade '%s'"), 3572 zhp->zpool_name)); 3573 return (0); 3574 } 3575 3576 void 3577 zfs_save_arguments(int argc, char **argv, char *string, int len) 3578 { 3579 (void) strlcpy(string, basename(argv[0]), len); 3580 for (int i = 1; i < argc; i++) { 3581 (void) strlcat(string, " ", len); 3582 (void) strlcat(string, argv[i], len); 3583 } 3584 } 3585 3586 int 3587 zpool_log_history(libzfs_handle_t *hdl, const char *message) 3588 { 3589 zfs_cmd_t zc = { 0 }; 3590 nvlist_t *args; 3591 int err; 3592 3593 args = fnvlist_alloc(); 3594 fnvlist_add_string(args, "message", message); 3595 err = zcmd_write_src_nvlist(hdl, &zc, args); 3596 if (err == 0) 3597 err = ioctl(hdl->libzfs_fd, ZFS_IOC_LOG_HISTORY, &zc); 3598 nvlist_free(args); 3599 zcmd_free_nvlists(&zc); 3600 return (err); 3601 } 3602 3603 /* 3604 * Perform ioctl to get some command history of a pool. 3605 * 3606 * 'buf' is the buffer to fill up to 'len' bytes. 'off' is the 3607 * logical offset of the history buffer to start reading from. 3608 * 3609 * Upon return, 'off' is the next logical offset to read from and 3610 * 'len' is the actual amount of bytes read into 'buf'. 3611 */ 3612 static int 3613 get_history(zpool_handle_t *zhp, char *buf, uint64_t *off, uint64_t *len) 3614 { 3615 zfs_cmd_t zc = { 0 }; 3616 libzfs_handle_t *hdl = zhp->zpool_hdl; 3617 3618 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3619 3620 zc.zc_history = (uint64_t)(uintptr_t)buf; 3621 zc.zc_history_len = *len; 3622 zc.zc_history_offset = *off; 3623 3624 if (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_HISTORY, &zc) != 0) { 3625 switch (errno) { 3626 case EPERM: 3627 return (zfs_error_fmt(hdl, EZFS_PERM, 3628 dgettext(TEXT_DOMAIN, 3629 "cannot show history for pool '%s'"), 3630 zhp->zpool_name)); 3631 case ENOENT: 3632 return (zfs_error_fmt(hdl, EZFS_NOHISTORY, 3633 dgettext(TEXT_DOMAIN, "cannot get history for pool " 3634 "'%s'"), zhp->zpool_name)); 3635 case ENOTSUP: 3636 return (zfs_error_fmt(hdl, EZFS_BADVERSION, 3637 dgettext(TEXT_DOMAIN, "cannot get history for pool " 3638 "'%s', pool must be upgraded"), zhp->zpool_name)); 3639 default: 3640 return (zpool_standard_error_fmt(hdl, errno, 3641 dgettext(TEXT_DOMAIN, 3642 "cannot get history for '%s'"), zhp->zpool_name)); 3643 } 3644 } 3645 3646 *len = zc.zc_history_len; 3647 *off = zc.zc_history_offset; 3648 3649 return (0); 3650 } 3651 3652 /* 3653 * Process the buffer of nvlists, unpacking and storing each nvlist record 3654 * into 'records'. 'leftover' is set to the number of bytes that weren't 3655 * processed as there wasn't a complete record. 3656 */ 3657 int 3658 zpool_history_unpack(char *buf, uint64_t bytes_read, uint64_t *leftover, 3659 nvlist_t ***records, uint_t *numrecords) 3660 { 3661 uint64_t reclen; 3662 nvlist_t *nv; 3663 int i; 3664 3665 while (bytes_read > sizeof (reclen)) { 3666 3667 /* get length of packed record (stored as little endian) */ 3668 for (i = 0, reclen = 0; i < sizeof (reclen); i++) 3669 reclen += (uint64_t)(((uchar_t *)buf)[i]) << (8*i); 3670 3671 if (bytes_read < sizeof (reclen) + reclen) 3672 break; 3673 3674 /* unpack record */ 3675 if (nvlist_unpack(buf + sizeof (reclen), reclen, &nv, 0) != 0) 3676 return (ENOMEM); 3677 bytes_read -= sizeof (reclen) + reclen; 3678 buf += sizeof (reclen) + reclen; 3679 3680 /* add record to nvlist array */ 3681 (*numrecords)++; 3682 if (ISP2(*numrecords + 1)) { 3683 *records = realloc(*records, 3684 *numrecords * 2 * sizeof (nvlist_t *)); 3685 } 3686 (*records)[*numrecords - 1] = nv; 3687 } 3688 3689 *leftover = bytes_read; 3690 return (0); 3691 } 3692 3693 #define HIS_BUF_LEN (128*1024) 3694 3695 /* 3696 * Retrieve the command history of a pool. 3697 */ 3698 int 3699 zpool_get_history(zpool_handle_t *zhp, nvlist_t **nvhisp) 3700 { 3701 char buf[HIS_BUF_LEN]; 3702 uint64_t off = 0; 3703 nvlist_t **records = NULL; 3704 uint_t numrecords = 0; 3705 int err, i; 3706 3707 do { 3708 uint64_t bytes_read = sizeof (buf); 3709 uint64_t leftover; 3710 3711 if ((err = get_history(zhp, buf, &off, &bytes_read)) != 0) 3712 break; 3713 3714 /* if nothing else was read in, we're at EOF, just return */ 3715 if (!bytes_read) 3716 break; 3717 3718 if ((err = zpool_history_unpack(buf, bytes_read, 3719 &leftover, &records, &numrecords)) != 0) 3720 break; 3721 off -= leftover; 3722 3723 /* CONSTCOND */ 3724 } while (1); 3725 3726 if (!err) { 3727 verify(nvlist_alloc(nvhisp, NV_UNIQUE_NAME, 0) == 0); 3728 verify(nvlist_add_nvlist_array(*nvhisp, ZPOOL_HIST_RECORD, 3729 records, numrecords) == 0); 3730 } 3731 for (i = 0; i < numrecords; i++) 3732 nvlist_free(records[i]); 3733 free(records); 3734 3735 return (err); 3736 } 3737 3738 void 3739 zpool_obj_to_path(zpool_handle_t *zhp, uint64_t dsobj, uint64_t obj, 3740 char *pathname, size_t len) 3741 { 3742 zfs_cmd_t zc = { 0 }; 3743 boolean_t mounted = B_FALSE; 3744 char *mntpnt = NULL; 3745 char dsname[MAXNAMELEN]; 3746 3747 if (dsobj == 0) { 3748 /* special case for the MOS */ 3749 (void) snprintf(pathname, len, "<metadata>:<0x%llx>", obj); 3750 return; 3751 } 3752 3753 /* get the dataset's name */ 3754 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3755 zc.zc_obj = dsobj; 3756 if (ioctl(zhp->zpool_hdl->libzfs_fd, 3757 ZFS_IOC_DSOBJ_TO_DSNAME, &zc) != 0) { 3758 /* just write out a path of two object numbers */ 3759 (void) snprintf(pathname, len, "<0x%llx>:<0x%llx>", 3760 dsobj, obj); 3761 return; 3762 } 3763 (void) strlcpy(dsname, zc.zc_value, sizeof (dsname)); 3764 3765 /* find out if the dataset is mounted */ 3766 mounted = is_mounted(zhp->zpool_hdl, dsname, &mntpnt); 3767 3768 /* get the corrupted object's path */ 3769 (void) strlcpy(zc.zc_name, dsname, sizeof (zc.zc_name)); 3770 zc.zc_obj = obj; 3771 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_OBJ_TO_PATH, 3772 &zc) == 0) { 3773 if (mounted) { 3774 (void) snprintf(pathname, len, "%s%s", mntpnt, 3775 zc.zc_value); 3776 } else { 3777 (void) snprintf(pathname, len, "%s:%s", 3778 dsname, zc.zc_value); 3779 } 3780 } else { 3781 (void) snprintf(pathname, len, "%s:<0x%llx>", dsname, obj); 3782 } 3783 free(mntpnt); 3784 } 3785 3786 /* 3787 * Read the EFI label from the config, if a label does not exist then 3788 * pass back the error to the caller. If the caller has passed a non-NULL 3789 * diskaddr argument then we set it to the starting address of the EFI 3790 * partition. 3791 */ 3792 static int 3793 read_efi_label(nvlist_t *config, diskaddr_t *sb) 3794 { 3795 char *path; 3796 int fd; 3797 char diskname[MAXPATHLEN]; 3798 int err = -1; 3799 3800 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PATH, &path) != 0) 3801 return (err); 3802 3803 (void) snprintf(diskname, sizeof (diskname), "%s%s", RDISK_ROOT, 3804 strrchr(path, '/')); 3805 if ((fd = open(diskname, O_RDONLY|O_NDELAY)) >= 0) { 3806 struct dk_gpt *vtoc; 3807 3808 if ((err = efi_alloc_and_read(fd, &vtoc)) >= 0) { 3809 if (sb != NULL) 3810 *sb = vtoc->efi_parts[0].p_start; 3811 efi_free(vtoc); 3812 } 3813 (void) close(fd); 3814 } 3815 return (err); 3816 } 3817 3818 /* 3819 * determine where a partition starts on a disk in the current 3820 * configuration 3821 */ 3822 static diskaddr_t 3823 find_start_block(nvlist_t *config) 3824 { 3825 nvlist_t **child; 3826 uint_t c, children; 3827 diskaddr_t sb = MAXOFFSET_T; 3828 uint64_t wholedisk; 3829 3830 if (nvlist_lookup_nvlist_array(config, 3831 ZPOOL_CONFIG_CHILDREN, &child, &children) != 0) { 3832 if (nvlist_lookup_uint64(config, 3833 ZPOOL_CONFIG_WHOLE_DISK, 3834 &wholedisk) != 0 || !wholedisk) { 3835 return (MAXOFFSET_T); 3836 } 3837 if (read_efi_label(config, &sb) < 0) 3838 sb = MAXOFFSET_T; 3839 return (sb); 3840 } 3841 3842 for (c = 0; c < children; c++) { 3843 sb = find_start_block(child[c]); 3844 if (sb != MAXOFFSET_T) { 3845 return (sb); 3846 } 3847 } 3848 return (MAXOFFSET_T); 3849 } 3850 3851 /* 3852 * Label an individual disk. The name provided is the short name, 3853 * stripped of any leading /dev path. 3854 */ 3855 int 3856 zpool_label_disk(libzfs_handle_t *hdl, zpool_handle_t *zhp, char *name) 3857 { 3858 char path[MAXPATHLEN]; 3859 struct dk_gpt *vtoc; 3860 int fd; 3861 size_t resv = EFI_MIN_RESV_SIZE; 3862 uint64_t slice_size; 3863 diskaddr_t start_block; 3864 char errbuf[1024]; 3865 3866 /* prepare an error message just in case */ 3867 (void) snprintf(errbuf, sizeof (errbuf), 3868 dgettext(TEXT_DOMAIN, "cannot label '%s'"), name); 3869 3870 if (zhp) { 3871 nvlist_t *nvroot; 3872 3873 if (zpool_is_bootable(zhp)) { 3874 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3875 "EFI labeled devices are not supported on root " 3876 "pools.")); 3877 return (zfs_error(hdl, EZFS_POOL_NOTSUP, errbuf)); 3878 } 3879 3880 verify(nvlist_lookup_nvlist(zhp->zpool_config, 3881 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); 3882 3883 if (zhp->zpool_start_block == 0) 3884 start_block = find_start_block(nvroot); 3885 else 3886 start_block = zhp->zpool_start_block; 3887 zhp->zpool_start_block = start_block; 3888 } else { 3889 /* new pool */ 3890 start_block = NEW_START_BLOCK; 3891 } 3892 3893 (void) snprintf(path, sizeof (path), "%s/%s%s", RDISK_ROOT, name, 3894 BACKUP_SLICE); 3895 3896 if ((fd = open(path, O_RDWR | O_NDELAY)) < 0) { 3897 /* 3898 * This shouldn't happen. We've long since verified that this 3899 * is a valid device. 3900 */ 3901 zfs_error_aux(hdl, 3902 dgettext(TEXT_DOMAIN, "unable to open device")); 3903 return (zfs_error(hdl, EZFS_OPENFAILED, errbuf)); 3904 } 3905 3906 if (efi_alloc_and_init(fd, EFI_NUMPAR, &vtoc) != 0) { 3907 /* 3908 * The only way this can fail is if we run out of memory, or we 3909 * were unable to read the disk's capacity 3910 */ 3911 if (errno == ENOMEM) 3912 (void) no_memory(hdl); 3913 3914 (void) close(fd); 3915 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3916 "unable to read disk capacity"), name); 3917 3918 return (zfs_error(hdl, EZFS_NOCAP, errbuf)); 3919 } 3920 3921 slice_size = vtoc->efi_last_u_lba + 1; 3922 slice_size -= EFI_MIN_RESV_SIZE; 3923 if (start_block == MAXOFFSET_T) 3924 start_block = NEW_START_BLOCK; 3925 slice_size -= start_block; 3926 3927 vtoc->efi_parts[0].p_start = start_block; 3928 vtoc->efi_parts[0].p_size = slice_size; 3929 3930 /* 3931 * Why we use V_USR: V_BACKUP confuses users, and is considered 3932 * disposable by some EFI utilities (since EFI doesn't have a backup 3933 * slice). V_UNASSIGNED is supposed to be used only for zero size 3934 * partitions, and efi_write() will fail if we use it. V_ROOT, V_BOOT, 3935 * etc. were all pretty specific. V_USR is as close to reality as we 3936 * can get, in the absence of V_OTHER. 3937 */ 3938 vtoc->efi_parts[0].p_tag = V_USR; 3939 (void) strcpy(vtoc->efi_parts[0].p_name, "zfs"); 3940 3941 vtoc->efi_parts[8].p_start = slice_size + start_block; 3942 vtoc->efi_parts[8].p_size = resv; 3943 vtoc->efi_parts[8].p_tag = V_RESERVED; 3944 3945 if (efi_write(fd, vtoc) != 0) { 3946 /* 3947 * Some block drivers (like pcata) may not support EFI 3948 * GPT labels. Print out a helpful error message dir- 3949 * ecting the user to manually label the disk and give 3950 * a specific slice. 3951 */ 3952 (void) close(fd); 3953 efi_free(vtoc); 3954 3955 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3956 "try using fdisk(1M) and then provide a specific slice")); 3957 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf)); 3958 } 3959 3960 (void) close(fd); 3961 efi_free(vtoc); 3962 return (0); 3963 } 3964 3965 static boolean_t 3966 supported_dump_vdev_type(libzfs_handle_t *hdl, nvlist_t *config, char *errbuf) 3967 { 3968 char *type; 3969 nvlist_t **child; 3970 uint_t children, c; 3971 3972 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_TYPE, &type) == 0); 3973 if (strcmp(type, VDEV_TYPE_FILE) == 0 || 3974 strcmp(type, VDEV_TYPE_HOLE) == 0 || 3975 strcmp(type, VDEV_TYPE_MISSING) == 0) { 3976 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3977 "vdev type '%s' is not supported"), type); 3978 (void) zfs_error(hdl, EZFS_VDEVNOTSUP, errbuf); 3979 return (B_FALSE); 3980 } 3981 if (nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_CHILDREN, 3982 &child, &children) == 0) { 3983 for (c = 0; c < children; c++) { 3984 if (!supported_dump_vdev_type(hdl, child[c], errbuf)) 3985 return (B_FALSE); 3986 } 3987 } 3988 return (B_TRUE); 3989 } 3990 3991 /* 3992 * Check if this zvol is allowable for use as a dump device; zero if 3993 * it is, > 0 if it isn't, < 0 if it isn't a zvol. 3994 * 3995 * Allowable storage configurations include mirrors, all raidz variants, and 3996 * pools with log, cache, and spare devices. Pools which are backed by files or 3997 * have missing/hole vdevs are not suitable. 3998 */ 3999 int 4000 zvol_check_dump_config(char *arg) 4001 { 4002 zpool_handle_t *zhp = NULL; 4003 nvlist_t *config, *nvroot; 4004 char *p, *volname; 4005 nvlist_t **top; 4006 uint_t toplevels; 4007 libzfs_handle_t *hdl; 4008 char errbuf[1024]; 4009 char poolname[ZPOOL_MAXNAMELEN]; 4010 int pathlen = strlen(ZVOL_FULL_DEV_DIR); 4011 int ret = 1; 4012 4013 if (strncmp(arg, ZVOL_FULL_DEV_DIR, pathlen)) { 4014 return (-1); 4015 } 4016 4017 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, 4018 "dump is not supported on device '%s'"), arg); 4019 4020 if ((hdl = libzfs_init()) == NULL) 4021 return (1); 4022 libzfs_print_on_error(hdl, B_TRUE); 4023 4024 volname = arg + pathlen; 4025 4026 /* check the configuration of the pool */ 4027 if ((p = strchr(volname, '/')) == NULL) { 4028 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 4029 "malformed dataset name")); 4030 (void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf); 4031 return (1); 4032 } else if (p - volname >= ZFS_MAXNAMELEN) { 4033 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 4034 "dataset name is too long")); 4035 (void) zfs_error(hdl, EZFS_NAMETOOLONG, errbuf); 4036 return (1); 4037 } else { 4038 (void) strncpy(poolname, volname, p - volname); 4039 poolname[p - volname] = '\0'; 4040 } 4041 4042 if ((zhp = zpool_open(hdl, poolname)) == NULL) { 4043 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 4044 "could not open pool '%s'"), poolname); 4045 (void) zfs_error(hdl, EZFS_OPENFAILED, errbuf); 4046 goto out; 4047 } 4048 config = zpool_get_config(zhp, NULL); 4049 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, 4050 &nvroot) != 0) { 4051 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 4052 "could not obtain vdev configuration for '%s'"), poolname); 4053 (void) zfs_error(hdl, EZFS_INVALCONFIG, errbuf); 4054 goto out; 4055 } 4056 4057 verify(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, 4058 &top, &toplevels) == 0); 4059 4060 if (!supported_dump_vdev_type(hdl, top[0], errbuf)) { 4061 goto out; 4062 } 4063 ret = 0; 4064 4065 out: 4066 if (zhp) 4067 zpool_close(zhp); 4068 libzfs_fini(hdl); 4069 return (ret); 4070 } 4071